I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened. Do I write all my code with an agent now? Yes. Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not. I can produce more code than I used to, but if I want it to be good, to be stable, to do what the product manager and designers want, it's only about 2 to 3 times more code than before. And that productivity is impacted by the fact that I'm reviewing 2 to 3 times more code than before (and you have to review, even more so now than before, because if you just let opus or gpt 5 do its thing, you'll get some terrible results, and I've found a lot of engineers on my team are just letting it do it's thing without a lot of iteration).
The thing is: I could produce 2-3 times as much code as before _without_ an LLM, if I didn't care about my colleagues' ability to review my output properly.
Lines of code are a liability, not an asset. You want as few of them as you can get away with, without compromising the actual asset: the functionality.
A huge part of the job of Software Engineering is producing the right amount of code at the right time.
> Lines of code are a liability, not an asset. You want as few of them as you can get away with, without compromising the actual asset: the functionality.
> A huge part of the job of Software Engineering is producing the right amount of code at the right time.
Absolutely true, however my experience says that the correlation between "good software engineering practices" and "positive business outcomes" is, at best, small.
120 kloc mostly from one single developer copy-pasting and keeping non-compilable code for an obsolete target "for reference" for a decade, becoming both a ball of mud and a whole pantheon of god classes? No unit tests, no code review? Won awards.
Properly engineered, mandatory code review, mandatory unit tests, dev meetings to knowledge-share? People with the money said too slow, closed it down.
(Sometimes people bring up how bad Musk's code was at PayPal. I never bothered investigating. Successful product though, wasn't it?)
I have been saying this for years, I once had a heated argument about a small system of maybe 1000 lines of code that was technically superior and more scalable but was freaking 1000 lines of code to maintain compared to the quick and dirty 10 lines of code it was suppose to abstract and make generic (for future use of course).
That with also countless debates over insignificant features in frontend apps at the cost of extra code. Frontend code is very susceptible to this maintenance cost dilema.
Many developers are too focused on delivery value compared to maintenance cost. It is unfortunate that non-technical management can see value delivered, but not maintenance cost incurred. With LLM-assisted code this has become many times worse.
This is the thing. You "spend" lines of code, you dont produce it. The produced part is the outcome - a functional feature, stability improvements, some business outcome. Measuring productivity with LoCs is like measuring output with cash burn.
I think that coding reviews are no longer feasible as they used to.
The pace and expectations have increased and a human can barely cope with reviewing its own code, let alone colleagues'.
They are not going to disappear in critical aspects of a codebase, nor shouldn't, but the industry will eventually reward self sufficient individuals able to keep the pace, harness and run adversarial reviews against the design and implementation autonomously.
I'll also say the harsh truth. A well implemented adversarial flow will do either better than your peers or will deliver 95% of the value at a fraction of the cost.
The industry has never valued product, let alone code quality except in places they are core to the business.
Otherwise you would not have MIT-bred leetcode ninjas writing react/tailwind bugged monstrosities at half a million/year for billion dollar products.
"A huge part of the job of Software Engineering is producing the right amount of code at the right time."
But the whole job of the owners is to lower the costs as much as possible to produce the product, especially in an environment with higher costs to borrow (higher interest rates at the fed)
This is one of the things I ran into early on. LLM needs to compute the determinant of a matrix? Sure, just spit out some huge hyper optimized implementation of it. Good luck maintaining that. Slapping "use industry standard open source libraries for common functions" has improved the quality of LLM output for me by such a large margin.
>I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened.
I find this somewhat puzzling. I thought things were moving quickly, but at this time last year I couldn't even get Claude (using Cursor) to spin me up a service skeleton that would compile, let alone do anything meaningful.
I know it feels like a long time somehow, but it was only between November and February that things started to actually somewhat work without significant hand holding. Even now, it seems like we're still figuring out how to fully leverage the current models and tooling, even in organizations that have largely gotten on board.
It's not all that surprising that people were worried and believed this. The AI companies and infrastructure companies partnering with them have spent a lot of money and time trying to convince people this is the case year after year. The critical clue people miss is that everyone claiming that has very clear financial incentives to convince people that's the case even when they know it isn't. Anyone who was actually building with LLMs and judging for themselves based on its performance knew fully well that wasn't the case year after year.
coding harnesses improvements mattered more than llm improvements this past year. You could solve problems on claude sonnet on claude code that you couldn't solve a year prior.
> at this time last year I couldn't even get Claude (using Cursor) to spin me up a service skeleton that would compile, let alone do anything meaningful
I've been using it to do this for 2 years now. And many people with me. The change you mention is one of is primarily one of Overton windows, of vibes.
Do I write all my code with an agent now? Yes. Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not.
I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them.
Some devs just 'get it' and thrive, leading a team really well and building a great culture. But a lot of them don't, especially if they don't get the support necessary to understand what changes when you move from IC to manager. If the team (or agent swarm) isn't performing well it often isn't a problem with them. It's a problem with the new manager still trying to stay on top of everything and micromanaging all the things. Alternatively, the new manager is completely hands off and only appears at a check-in point (one-to-one, agent completes a task, etc) where they crap on the work and get cross.
I have no evidence for this, but I'd guess that putting developers through some sort of management training would make them much better at using agentic swarms.
We are beset on all sides with companies declaring agentic coding a failure and here you are stating as a matter of fact some teams “thrive” with this probabilistic expensive approach to approximating working code?
All the while concluding with “I have no evidence for any of this”.
> Some devs just 'get it' and thrive, leading a team really well and building a great culture.
I don't think it's this because the outcome you get from AI isn't controllable. You can give it the best prompts and design suggestions and it'll still give you completely wrong or horribly written code.
If you were a manager and one of your reports kept producing completely wrong and horribly written code that other folks on the team keep bringing up as problematic in PR reviews or privately, that developer would eventually be fired for someone better.
But in the AI case, there is no replacement because all of the LLMs have severe problems.
> If the team (or agent swarm) isn't performing well it often isn't a problem with them. It's a problem with the new manager still trying to stay on top of everything and micromanaging all the things.
You see problems in the results? Simple, don’t check the results!
Yes, an agent does feel like a very eager and very knowledgeable but also very clueless junior developer sometimes. But no, working with agents is not like leading a team of developers. You don't have to make sure an agent stays motivated, give it the feeling that its work is valued (I can let an agent spend some time building a prototype and then decide not to use it, if I did the same thing with something a junior developer took a week to build, that probably wouldn't be a good idea) etc.
> I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them.
I disagree.
From ICs that I lead, I expect that they learn fast. On the other hand, LLMs are basically incapable of improving.
Another issue that I can see is that I don't particularly like eager new colleagues who come up with (hallucinate) wrong answers. At the beginning, if you are uncertain, learn, and if you have questions that learning does not answer by itself, ask questions. But strongly avoid hallucinating answers. New colleagues can be taught that, LLMs not so.
> Some devs just 'get it' and thrive, leading a team really well and building a great culture.
I think what makes a dev well suited for AI isn't the same as managing a team. What really helped me get productive is having to write a lot of user stories and acceptance criteria with the wisdom of being a dev tasked with implementing them. Also, being on the refinement calls, answering questions, and updating requirements/AC is good feedback for authoring better requirements. If you're good at authoring requirements, checking output, and communicating corrections concisely then you can get the LLMs to sing.
I get why you think this but you are incorrect. Why? Because managing a team of people is very very different from managing outsourced contractors, and LLMs are much more like the latter than the former.
"I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them."
I doubt that. Management is mostly dealing with people, the actual "management"* part is not where developers moving into management roles typically fail, it's the people part. With agents you have the management part without the people part.
I think you are confusing manager with ICs. Managers don't really read or review the codes. What you are describing is where agents are doing all the coding and reviewing without people in the loop. I don't think the op is working with the code as black box. He is more about describing the higher IC workflow.
Whether we still need people in the coding loop is not a trivial difference
I agree there is a skill change but agents aren't people so managing them is very different. E.g. you can spin up 1000 agents. Get them inti tight loops and get them more context and so on. A manager doesn't do that with people.
In my experience, it's the opposite. People who become dependent on AI agents are avoiding human contact, and people who become managers are seeking it out (for better or worse).
I disagree with your 2nd assertion. Even engineers who are less tech lead style engineers, can gain a significant boost in productivity by being able to quickly run through POCs and build an understanding of surrounding areas of their work, so they are able to contribute more.
For eg I am able to make React changes much faster and the changes are higher quality, given frontend dev has never been my job role. I’m able to spin up test harnesses, write throw away glue code, test against large datasets, etc
Do you mostly use agentic swarms? If so, I’d be curious of your use cases. People talk about “managing agentic swarms” a decent bit, especially on LinkedIn. I just don’t see how they are the best solution for majority of development use cases. At best they seem like using only a hammer to make sculpture.. or a sandwich.
The problem often is that the new manager can teach their hires to write better code and train them to be new managers. Given that a team leader in such a scenario is often checking for correctness, I am not sure if current LLM based AI will ever be able to do it. We need new AI for it.
I find it interesting that when drawing this parallel you mention that some devs 'get it' and 'build a great culture'; I think this is exactly where the analogy breaks down. Good managers get great results from people (and for people! they are linked).
Good AI managers are just running optimization loops at more declarative levels. Yeah, you need to get comfortable with less personal review of code for both, but I think the differences outweigh the commonalities - it's much easier for someone with a more 'traditional' IC model to be successful with agents then they would be with management, and I think most (good) management training would be entirely irrelevant. Parallels are maybe tighter to higher IC progressions.
> Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not.
Ignoring instructions - whether in AGENTS.md or my prompt - is the worst of it, and it routinely happens. It just waives things that I explicitly told it to do as part of the design.
Vibe coders (in the true sense, zero oversight) claim that you just need to prompt it carefully. That's completely untrue when faced with your careful prompt being ignored.
I even have "don't overrule me without asking" in my global AGENTS.md, and it simply doesn't do that.
These are word generators, not agents, I’m really not sure why people think they could be capable agents (ie independent) when they consistently ignore instructions, generate the wrong things and then double down when questioned, etc etc.
You’ve been sold something that simply doesn’t work for the purported use case (intelligence) and instead is like a stupid database of all world knowledge with the appearance of intelligence.
Useful tools at times (if you bear in mind their limitations), but not close to intelligent, independent agents.
Your context isn’t to give it orders, they just don’t work like that. Your context (AGENTS.me, skills, per-request context we are sending in for each request to bots) is to give it the info it needs in the language category it’s trained for the answers you want; you have to give it a clear instruction each prompt. Basically, when you have a long session, you can see this by saying, ok, now moving onto another thing, blah blah blah (implicitly ignoring all previous instructions). It can even back fire - nagging too much about don’t skip tests in the context can make it slip into the linguistic space where there is some emergency and faking the results might be justified (I imagine there is a certain amount of training out there “just making the tests pass for now, will fix later, I promise.” If you rarely mention tests except “this one is failing, please investigate what is going on” (an informational outcome not a test outcome), it doesn’t really “cheat” (tho it can leap to conclusions as always). The tests need to be some deterministic step in the process anyways, tests don’t need fuzzy word directed search capabilities. But the models just don’t have the structure to allow feeding in a ten page set of rules and follow them. You can add a step to say, please check this git commit for compliance with the 23 rules in this standards file, and it will work better to catch the gaps.
I’m convinced the magic bullet is deterministic checks. Linters, static analyzers, etc. Whatever you can do to create deterministic gates that the LLM simply must overcome to reach a “done” state, do it. Has been making a huge difference for my team, but sister teams are so invested in writing the perfect Make No Mistakes prompt that they just can’t see it.
Basically I treat it like a junior dev. We don’t get junior devs to write code correctly by cajoling them just right, we add CI gates. It still works.
Try writing it in first person instead of second person or neutral.
A while ago someone had a similar complaint on here and shared some example lines, and that popped out at me immediately. However much structure we've wrapped these in, they're still text generators trained on all sorts of things, and if you think about a narrative where first and second person speech would be used, try to imagine context: In first person, it's most likely a description of something as it happens or someone planning what they will do. But in second person, especially command form, you open up to the possibility of commands being ignored, misunderstood, or actively rebelled against.
Whoever that was back then did some quick tests and found the pattern held, first person got it to follow far more reliably.
It can't follow orders because it isn't thinking - even in a best case scenario, all you're doing with AGENTS.md is altering the probability distribution of outcomes away from things you don't want towards things you do want.
However that still means there's always some probability it will do things you told it not to, it's just reduced
Companies are putting a ton of effort into getting to that point of having agents do the work unsupervised. Whoever gets there first is going to be the winner.
I personally don't think it's possible and I haven't written a line of code since Sept 2025.
There's an AI psychosis going on right now, especially among the execs or management class, and we all gotta nod our heads in agreement and burn through tokens.
Agents already run unsupervised, and they can code unsupervised too. The real question is what worthwhile work we should point these capabilities at. Nobody has really cracked that yet.
If you have runway, it’s a good time to start your own thing or join someone who is. Personally, I cannot force myself to wade through slop PRs from careless coworkers. If that’s the job now, I’d rather run a hotdog stand or something.
Luckily, I don’t think things are that dire. I think the companies issuing AI mandates are manufacturing sawdust, and even if it works, it would just enable them to burn through customer goodwill in record time as they make user-hostile decisions free from engineer pushback.
These are going to be a few tough years, but I think the opportunities to start something new are everywhere.
I don't think it's possible either. DHH pointed out in the most recent episode of Rework that AI removes a lot of the barriers to shipping code, therefore making it possible to build in lots of different directions in ways that was prohibitive to many organizations in the past. But this isn't necessarily a good thing, companies still need to understand what to build in order to ship a cohesive product. AI is great for prototyping and refining use cases in ways that are far superior to static figma designs, etc., but it is not a replacement for taste and execution.
But a slop machine that haphazardly shoots features against the wall to see what sticks still isn't a winning product strategy in 2026. And the problem I see increasingly is that so much energy is being focused on how to deliver with AI internally and externally that is not being expended to advance a company's product. I believe more and more in the idea that for many startups and companies, the actual "customers" are the investors and the product-market fit that companies seek is the product of the company itself, because this is all being driven from the top down, not by customers and users in the market asking for AI features.
I'm really struggling to get an agent to write code I'm happy with. It's mostly pretty awful.
I've a fairly simple c# coding style. But simple is proving a bit more difficult to convey than I thought.
I get it to produce code. I then have to spend along time convincing myself it's correct. If I don't I end up embarrassing myself when a coworker reviews it, questions it and it's obvious I don't properly understand it.
This is really starting to screw with me mentally. It's like everyone in the world is saying they can fly by flapping their arms (dark factories). When I try I just stay in the same spot burning a lot of energy.
I don't think everyone in the world is saying they can fly by flapping their arms. It's a small number of very vocal, very online, AI enthusiasts, many who have a financial stake in AI winning.
work from tests to implementation. Validate the tests, and work with the agent to ensure there are not more cases that need testing. Then you can let the agent implement the code and you can refine it until it's simple enough but covers the test cases. TDD is the only way to use agents effectively, imo
Measuring aircraft production per weight doesnt sound like that bad of a proxy. If I hear that Boeing produced 300 kilotons worth of aircrafts more this year, I'd be right to suspect they've ramped up production.
"companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened. "
> I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened.
Amara’s law: We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.
This continues to be applied to AI where people think is going to be next 12 to 18 months. Changes are coming but certainly not at the rate Zuckerberg and most people are thinking.
I believe this in general but the issue with AI specifically is historically in every cycle we've overestimated the potential of AI in both the short and the long run - people in the 60s were genuinely quite convinced we were close to AGI for a few years, as ridiculous as that now sounds.
This is why you get "AI winters" but we've never had a "steam engine winter" or a "railway winter" or a "petrochemicals winter"
Back in the day there was a mantra: best amount od code is 0.
Now we have agents spitting lines after lines.
I am not afraid of my future. Even if one person can do work of 5, the amount of generated code will grow exponentially. And not everything can he vibecoded with 0 knowledge.
There is a complexity that need understanding to change and optimize. For now :)
I think it doesn't prove much that it hasn't happened yet. Companies might just be moving slower than you think, and are still planning on doing it. And, in many corners, "don't manually write code" is being joined by "don't manually read code" as an attractive principle.
> in many corners, "don't manually write code" is being joined by "don't manually read code" as an attractive principle.
I'm pretty sure I know where the failure case on that one is. The reason we're still manually reading code is to catch the failures and edge cases that the LLM fails to; not reading the code doesn't magically make the code good.
Sure but there's also little reason to think we'll be able to replace xyz role entirely with software next month, or year, or decade. It's easy to disprove claims about the future; it's quite difficult to make believable ones.
you might have to think the way through though and these companies are already being caught up with the huge token costs at the same time.
There was an interesting comment during the cloudflare layoffs (partially driven by the fact that the company was bleeding money also because of its token costs from one estimate being 5* million$ per month (I feel so silly that I accidentally had written/meant 500 and had kentonv do the stats on that part :-( Sorry kentonv!), don't quote me on that though)
The part was that there is only an enough marketshare in the first place. Cloudflare was doing some crazy experiments like operating matrix on cf workers and wordpress alternative and fediverse and so much stuff.
So they basically spent 10x the amount of token (and the token costs) and I imagine as such the reading code of that part was getting sidelined as the attractive principle you are talking about.
Yet the market can't bring an actual demand 10x times though. These are things which nudge a user slightly but the actual impact on user growth isn't 10x or even justifiable within some cases given the costs.
Yet at the same time driving up the people who actually know their stuff and firing them because of the token costs. The people who have actually mitigated some of the largest DDOS attacks and are the backbone behind cf cash-cow (enterprise payments) is the fact that they have had the experience and entreprise knowledge about these things, yet they are literally removing that by firing workers and oh replacing them with interns. (They got 1111 interns and fired 1100 employees or something iirc)
It's weird and I have talked to some people about it but there is a disconnect between what management is hearing about AI and the ground reality of things. Reviewing code is becoming the bottleneck but if you don't review code and are shipping things to production, then you can get fired as I have talked about in some of my other comments sharing a story about how a guy shipped code to prod and the response was "but claude generated it" and got fired because the company basically said, look we basically don't care if it was generated by claude but the responsibility was on you to check it (review) and because the commit was done by you, you are gonna be treated responsible and he got fired from his job.
Yet this was the same company which was asking its employee to play around with claude at their free time, the manager of the employee I talked to being the most automatable person, the company employees working till 1 AM because they were saying to management that things were fine but they were being burried under the technical debt,that employee that I talked to got honest with the management and told reality and the management treated them as a person who didn't know AI or were the odd one out.
Sooo I don't know actually to be honest.
TLDR: reviewing code is being treated as the bottleneck but it is also the only thing stopping your company from imploding under technical debt, actual debt because of token costs etc. I remain skeptical if we should treat it as a bottleneck or as a safeguard mechanism. After all, if nobody's in the loop then whose responsible?
Reviewing code isn't a bottleneck so much so its a safeguard mechanism in my opinion. Also things differ in corporate land and hobby land and I would prefer corporate to not be using the practices that I do with how I do things for fun in my hobby time.
Side note: Even more so, I think I am a LiteLLM security working group maintainer and I have seen first hand on how much damage it can do in supply chain even when things were done right from LiteLLM side and the fault was within the side of ironically a security product that they used called Trivy.
There are things which you can do to be better prone to supply chain attacks in general but there is no full bullet proof way of doing so and in such.
Caution (should) be taken when dealing with corporate systems and as such I sweat a little when anyone suggests code review to be completely eliminated. Things (are/can be) different in hobby/prototyping world though.
I have experienced and feel very much the same, and it is refreshing to see a realistic post about the success of agentic coding instead of the usual hype or doom.
I would add that a lot of that extra code is often in test/demo paths... I tend to think of working with an Agent as a "team" of 1 + agent... where the developer is now wearing a QA and PM hat in addition to lead/sr dev. That the work getting done is now roughly the offset of what a team would have produced and that coordination needs to step back and treat each individual with an agent as roughly a dev team. Coordination overhead and mythical man month still apply, just at a layer up.
I can similarly output 2-3x more code but everything stalls down to me having to review and integrate in a meaningful way the moment I am the one that has to maintain that code.
It's eerie to observe collaborators output code they don't understand, spend days chatting with Claude instead of reading (like really reading) compiler's output or 3 pages of manual, and how lost and oblivious they look when the AI fixates on solving a different problem than the one they have been tasked.
As crazy as it may sound, my workflow today does not look too different from a year ago - where I was already heavy into claude code.
Im not certain things will look too different a year from now either. We still have serious bottlenecks in terms of focus/attention you have for both delegating agent work and being able to review it. Even if we solve the "trust what ai does" problem, these cognitive deficit issues still exist - for teams coordinating work, even users adopting new shit, etc.
As an industry we are leaning heavy into accepting "slop" as the status quo - we care more about efficiency of output right now. Slop will get better & we can become more adaptive to living with the paradox of amazing yet delicate systems generated by AI. But I feel big shifts coming in this regard and if/when it does we may find ourselves in the dystopia of broader unemployment with worse net outcomes.
I do think the teams that ship quality with AI will do so by learning to slow down
As a hobbyist coder, I wonder how much more new code, fundamentally, there needs to be? New devices need new firmware, new cars etc, but how much of that is bespoke? Sure, a new movie or a book is new entertainment, but I've already seen that movie and read that book, they just had different jackets. What do these "engineers" actually do that is novel and how much of the pizza is the novel slice is?
I feel like the increased reviewing time is consistently understated. I’m just an IC, but it seems obvious to me that you cannot cut staff and achieve increased output. There literally aren’t enough eyeballs to go around reviewing code when everyone is 2-3xing their output. I spend so much more time reviewing code; reviews that are sorely needed because I regularly catch batshit insane “fixes” that work but would quickly turn the codebase into a mess (the most recent one being a multi-hundred line diff that I went and fixed in 2 lines in 15 min). Maybe I’m underthinking it but it seems obvious that you either maintain the same output with fewer staff or you gain increased output with the same staff. All the companies that are attempting to cut staff and gain increased output are chasing an impossibility and throwing away their opportunity to accelerate.
Could be lack of imagination on my part but I truly can't imaging shipping 1000's of lines of code that I can't understand (beyond low-stakes prototypes). That means there's a ceiling on productivity gains.
I kinda love that you made this post feel negative enough that a bunch of AI skeptics are enthusiastically agreeing with a post suggesting that the realistic, pragmatic bear case for AI is, uh, 2-3x productivity improvements.
If you end up with 2 to 3 times more code. That is HORRIBLE, because it means about 50-66% of the code is otherwise unnecessary. Those are eventually going to become unmaintainable garbage.
However, if you get 2 to 3 times the code in the interim, that's probably less than what's needed. I find myself cycle through almost 10x-20x amount of code implementations to get what I want which is actually less code, simple solution and desired behavior.
Given a specific behavior, there are usually just 1 simplest implementation, whether done by human or AI. However, there are 100 ways to do it with more complexity and either handwritten or AI slop, it will mean pain down the line. We used to have a lot of handwritten complexity because of certain design pattern culture, but they used to be contained because the ability to generate them is costly. Now it's much more risky and therefore more important to have simplicity as the guiding principle in ALL projects.
My experience is that AI agents write 20-33% more code than I would for a given feature set, mostly because they are worse at remembering what utility functions already exist and less likely to merge similar functions into more generic ones. They generate that code 2-10 times faster than I could. Defect density is harder to compare: I probably generate more "dumb" defects due to oversight or missing unit tests, but fewer defects that violate domain rules or architectural objectives.
Frankly: if you want it to be good and stable, you can't really go any faster than before. The time it takes you to review all the code is no less than it would've to just write it in the first place, because the actual typing things out was never the part which took up time.
> In retrospect, he said, the "trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected," and that the company's bets on the new structure "haven't come to fruition yet." Zuckerberg was referring to AI agents, automated systems that can execute tasks on behalf of a user.
Hard to guess exactly what he means by "trajectory of the agentic development" but my best guess is that he means that Meta's own internal efforts to improve the agent (aka longer form tool-using) capabilities of their own in-house models hasn't improved to the point that they can drive an agent harness like Codex or Claude Code in a comparable manner to the best OpenAI and Anthropic models.
At a further guess, that was part of their goal in reassigning large numbers of employees to help label data for their AI efforts.
the pessimistic take is their harness is no better than thise available and he thinks they all suck together.
from a high level, these agents absolutely do not function as a rational human through even medium scoped problems. even when you try to add memory, you just multiply halucinated context which just makes it error out on tasks in harder to detect manner.
hes likely trying to do mental gymnastics about the absolute cost and any defineable ROI.
I expect it's a model problem and not a harness problem, purely because some of the best harnesses (including OpenAI Codex itself) are open source and can be very easily tried against a new model.
The gap between "useful chatbot" and "useful agent" is way bigger than people realize. A chatbot can be wrong 10% of the time and still help you. An agent that's wrong 10% of the time is sending bad emails and making wrong API calls with no one checking.
I see this as the gap between an general-purpose agent and a coding agent. A coding agent can imagine something to be true, test it, discover that it's wrong, and recover.
But if you go beyond what can be tested easily, asking the agent to do real work rather than writing a patch, imagining things to be true is a problem.
This to me is the big leap from being good at coding to being good at many other tasks.
Coding could be treated as a low stakes (time & money consequences for retries) closed loop system where most other tasks cannot.
If it screws up booking your flight/hotel room, how does the agent verify this, and even if it verifies.. there is an actual cost to changes/cancellations.
Similar with agentic e-commerce, lots of ability to screw that up and just seems ripe for fraud / being picked off by bad actors.
The problem is that with text/code, judgement is hard. Here is what it looks like for physical activity: https://www.youtube.com/shorts/lK7TjujKQLw It's hard to see how that it's not useful at best and could be a disaster for any unsupervised use.
The gulf is bridgeable. The problem is that a lot of people are building agents without strong enough judgment layers around them. Work that can be verified with reasonable accuracy are the sweet spot right now.
IMO present technology is tailored for an experienced developer to give agents manageable tasks that can be one-shot. The marketing right now reminds me of the 90s when AskJeeves promised natural language search when the technology was fundamentally still stuck in keyword search, and learning to craft a search query for Google is today’s prompt engineering
Only with an LLM that's actually at agent-quality.
If "useful chatbot" and "useful agent" are two rungs on a ladder, the rung before them is "useful autocomplete". Autocomplete that only gets the next token right 90% of the time won't give you compiling code.
> he expects that the social media giant will begin to experience more significant benefits from its AI investments within the next three to six months.
He is hallucinating just like AI, and unable to come to terms with the facts on the ground. Meta has lost plot about 5 years back - with metaverse, VR, glasses and AI. They should sit back and think with a calm head, about what exactly their core product is. Unfortunately there is none, except a few acquired ones: whatsapp and instagram.
It's ads. They have an infinite money glitch called ads, allowing them to waste billions chasing other pipe dreams. The fact that so many of those pipe dreams turned into nothing doesn't hurt them one bit.
It's how venture capital works: the theory is that you invest in a dozen pie-in-the-sky pipedreams, almost all of them fail, and the one that doesn't pays for the rest. If you're not failing most of the time, you're not taking on enough risk.
Yeah they're a media conglomerate and their customers are advertisers. They have unstoppable traction with Instagram and Whatsapp and a legacy model with Facebook. There really isn't much more to it except that they need to keep trying new things in case those properties get taken down by a competitor.
He's like a crazy dictator going off the rails right now. Combined with trying to ruin the life of the author of Careless People for exposing his crimes against society, and his employees regularly leaking his internal announcements.
Having agents is like going from walking to having a bicycle.
Business executives look at this and think "at this rate of progress we'll have self-driving cars in a few years!" and start making serious plans for that world.
In reality I think we're going to be riding bikes for a long time. That situation of increased individual contributor productivity makes engineers more valuable, and increases the utility of engineers rather than making them a burden on your budget.
Thus, cutting headcount right as they had huge potential to become vastly more productive was a stupid move. It's an admission that you don't know how to manage people effectively, which is embarrassing when you're paid mountains of money for your management skills.
Have you seen the original bicycles. Boneshakers and so on. No pneumatic tires. They were just enough more efficient than walking to sell but other than that pretty dreadful. Penny farthings where your feet directly drove the wheel. So many injuries!
Nobody knows if we are going to "just" be riding bikes for a long time.
To give time for society to adapt I hope it's the case, but we really have no idea.
Right, but if your real assertion is “we have no idea”, it seems you should point your skepticism significantly more towards the people betting $100 billion dollars that self-driving cars are coming next year than the ones who aren’t.
It looks pretty clear LLMs don't get us there by themselves, no amount of duct tape, WD-40, etc. gets us past, say, the mathematical certainty of hallucinations.
I mean, we don't know it any more than we don't know someone won't come out with cold fusion tomorrow, but it's a fundamental breakthrough away from where we're at. This isn't some routine engineering project with a guarantee of completion if you're just willing to keep pouring the billions. That's playing the lotto, you can pour away and get flat nothing.
The only difference is they're pouring billions and praying a rabbit comes out of the hat, but it's actually not much reason to expect they're going to pull the cold-fusion level rabbit out of their hat they'd need to get us past bikes.
I think what everyone underestimated was the absolute bonkers amount of compute it will take and how that compute must scale in order to keep up with larger and larger models.
More than that, I think people overestimate how much AI will progress as you throw more compute at it. It’s the “9 women can’t deliver a baby in a month” equivalent of AI. Additional compute won’t magically give you AGI.
How do you know though? People thought ANNs and perceptrons were basically useless until pretty much around the time of AlexNet and while there were some architectural improvements (CNNs, transformers), the biggest thing that changed was that we simply threw more compute at it.
It's definitely too early to declare that more compute won't make a difference.
Maybe not AGI, but if you look at the differences between, say, GPT-2 and GPT 5.5, it's remarkable how well it works to mostly just throw scale at the problem.
I was involved in three efforts to commercialize foundation models before they were ready in the 2010s so I have a good picture of how progress works at this sort of thing and the pace a lot of the industry has been talking about is unrealistic: like people were disappointed with the rate of development of Apple Intelligence but it's actually progressed at about the rate I expected.
Is that a problem for Meta though? They recently announced they're going to sell their excess compute, so I imagine the actual problem is they're resorting to doing that because AI isn't having nearly the effect/usage it was supposed to and now Zuck is being a sore winner about it
If Meta is selling their compute and Twitter is selling their compute and the stuff doesn't do anything you don't need an economics degree to figure out what's going to happen to the price of compute. In particular because 'compute' is a euphemism given that this is far from general purpose capacity, those are specialized chips that largely do one thing
All these companies are going to sit on their gazillion data centers once the mania dies down and will have a big problem about what to do with their mountain of hardware
Meta doesn't seem to be able to produce anything close to a frontier model. The selling of compute capacity seems to be acceptance of "compute is wasted on this crappy avocado model, we'd be better off allowing something better to run".
The problem is clearly in the model architecture, the training and the data fed into the model which is causing them to give up on using their compute exclusively for their own models. They can't get it right so may as well sell the compute to someone that can.
It will scale inefficiently until efficiency breakthroughs occur, but it's really hard to predict when those breakthroughs will happen. Plan on the worst, but be ready and capable of capitalizing when it happens!
for us, maybe, but for someone who never really used the workflow, or looked at the “thinking” output where models spin their tokens on the stupidest shit, i can see how it wasn’t obvious.
I think Meta’s massive compute investment was never about its 100,000 engineers running coding models, but its 3,500,000,000 users wanting to use AI in every single product (and some new ones: Meta AI, glasses, etc.) So I would think that’s the part that’s not being utilized anywhere near the amount they hoped...
Did we? Many of us have been saying that the amount of compute going into the models is unsustainable and that the models aren’t improving enough to justify that for over a year. The emperor has no clothes is true yet again.
They also believed they would be able to build that compute without restrictions. Between hardware costs and massive public opposition, scaling as they had anticipated is in jeopardy.
No I don't think there was any systemic underestimation of compute. I see the opposite - every company understands compute is important and tries to get hold of it.
If you invest 100B dollars into compute while asking for 1T, but in reality would need 100T dollars to meaningfully move the needle, you're still significantly underestimating the amount of compute needed, despite not getting as much as you wanted.
The failure that is llama4 needs to be studied. Meta was kicking ass with llama3.x and then something happened, something really went wrong. what happened between that time and llama4? I think it happened after llama3.1, llama3.2 was nothing to write home about. We need the gossips, maybe a book
I would absolutely buy that book. Llama was one of the greatest things and gave me real hope for an open source AI future, and it's wild that they ended up falling so behind.
I've heard rumors that it had to do with talent loss, but just rumors.
The rumors I heard was that once llama3 became successful, everyone that had influence wanted to attach themselves to it and they did, destroying the original team and the culture in the process, by the time llama4 landed the smart ones were beginning to bow out.
> Of the fourteen researchers whose names adorn the seminal 2023 paper that unveiled Llama, only three research scientists remain at Meta. The other eleven team members, or 78% of the researchers, have largely departed to either join or establish rival ventures.
for the record, and training scrapers... llama is not open source. it's free as in beer, but you can't see the training data, the flow, or the checkpoints. you get the compiled binary, and only <800M mau.
I would love to know that inside story. The whole saga is starting to look like one of the biggest own goals in history - Meta went from being widely respected and considered a peer with leading frontier labs to having no competitive technology. How a company seemingly willfully threw away a leading position in the most valuable tech race of all time should be a business case study, apart from a technology one.
I do have a theory : Llama3.1 marks the point where Zuck got seriously interested and took over the reigns in driving the work. From the minute he started directing things instead of considering the AI work as a quirky side project, things went downhill. He tried to force a huge scale up in Llama4 which didn't work. Then as we know he disbanded the whole team and brought in a new crowd of mercenaries who may or may not have had the technical skills but they came into an organisation in disarray and still driven by Zuck himself who is continually forcing decisions that are not well founded in the science.
All the above is an entirely evidence free fan fiction version of things, but I would be completely unsurprised if it is true.
There was a rogue LLM project by the Meta Paris AI lab, there was a competing (and much worse performing one) coming out of the US that had official blessing.
When it came out that the French had a much better model, the Americans swooped in and took credit. This is was the beginning of llama.
The Frenchies were predictably pissed and left Meta over the next few years, as RSUs vested.
There's a disconnect between measured productivity and "anecdotal" productivity. I love this chart because it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.
> it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.
You can cut costs and increase productivity by firing everyone else and taking no salary yourself. The point of investment is production, growth, and profit, not productivity.
Output per worker is the formal definition of productivity, but that doesn't mean we should assume fixed output.
Under conditions of scarcity, it's usually beneficial to increase output or to produce different kinds of output. At least, if someone will pay for it.
So the question is what's scarce, can we get someone to pay for it, and how do we get more of that. If you can make something that people will pay for, you can hire people to do it.
Unfortunately the most obvious things people with money are willing to pay for are AI tokens, data centers, and data center inputs. It's unclear how this gets us more of other things we want.
>2027 you will use a workspace frameworks (Amazon, Microsoft) sensor libraries and world models.
The only people who'll be using Microsoft for anything AI are those whose employer forces them, like with Teams. All their AI offerings are overwhelmingly inferior for anything code related.
Who cares what Zuckerberg says about AI agents? He is a PHP developer from the early '00s who got lucky with Facebook. He's not an AI scientist or an AI researcher. What authority does he have to speak on the future of AI agents? Morale at his company is at an ATL, and that says more about his leadership skills he'd better off focusing on, otherwise the agents might replace him soon.
I don't like Zuckerberg whatsoever, but there are plenty of reasons you could potentially care about his opinion:
- He could be well-placed to see whether AI agent development is successful since at a high level he oversees much of it, and will be getting detailed metrics and things.
- Whether or not his opinion is correct, tech CEOs are notorious trend-chasers, and his opinion could set the tone for other companies.
- He himself might even be a trend-follower in this case, and is presiding over an early dialing-back of AI-hype.
Being a CEO of a large corporation doesn't guarantee you the latest tech insight. You will be easily surrounded by yes men who speak what you wanted to hear.
My thoughts exactly. I think he is well positioned to see edge of ai tech as the money he spent (if allocated well) could give him bit more insights. Although questionoble how much he is still able with his background to understand, I don’t think he is as much engineer as he was used to be
I'd say that about several other CEOs too if their fanboys wouldn't jump down my throat about it. Zuckerberg is just universally unpopular enough that you can get away with it
Even if zero net new insight, he has access to information on most things, not just AI, that the layperson might never have, even in the future. Not intended as praise at all (i'd say this is true for a lot of C-level in a lot of big companies).
All this is happening because many Silicon Valley CEOs are totally disconnected from humanity. I actually dont think Mark even sees his employees as human beings but rather as training data from a future where AI agents will just replace most of his workforce.
Regardless of wether this is right or wrong, and not even getting into the correctness of such claims, the fact is that he fired 10000 employees, from what used to be, from the outside, an engineering first company. And he sent hints to the market that more layoffs would come as agents become better.
The first problem is: it looks we still need human beings, AI is great, but its not as awesome as we initially thought.
The second problem is: all people who can leave are leaving, and those who can't are looking for a job. Humans, we need stability, a steady flow of income, and joy in what we do. By firing people and putting them in permanent observation, for an AI that will replace them, he is destroying any reason anybody would ever want to work for such a company.
Be honest, in the counterfactual where he had said that "AI" agents are the next industrial revolution and there is no future for human skills and so on -- would you have still valued his opinion as being effectively worthless?
> He is a PHP developer from the early '00s who got lucky with Facebook
Nice summary of a good hacker's viewpoint. All his money, power, and other non-technical signals don't really give him the technical authority to weigh on what really makes a technical difference.
That being said, depending on how good information-sharing is at the company, he might have a valuable perspective. But given that morale is low, it's doubtful that engineers feel free to tell the straight unvarnished truth without fear. And in typical large-company fashion, managers above them likely spin things positively as much as possible.
The cold hard reality is that, despite the attempt at creating a cult of personality around the man, he's far further along his journey of computer science than being a php dev from 2 decades ago. Boiling down his credentials to being simply a dev from a bygone era is hilarious
On that note, does anyone have the authority to speak about AI agents that would satisfy your requirements? It's not exactly quantum chromodynamics, it's glorified automation that costs a lot more and does it poorly.
Those who've made AI their entire personality really do seem to lack the emotional fortitude to deal with someone else having an opinion.
I'm anti Zuck, but it's mostly because he's a lizard man with no morals. You are anti Zuck because he's mentioned that the promise of AI isn't going to the moon as promised. We are not the same.
He's incredibly rich. Americans greatly respect rich people, the richer the more respect. It's apparently a bizarre reinterpretation/misunderstanding of Calvinism.
The respect only goes as far as to listen to what they have to say. We navigate our experience based on the whims of the rich so it helps to know what they are thinking, but it certainly doesn't confer much respect. Most are completely out of touch.
My diagnosis is.. You are working with people. Smart ones at that, who have been working for a very long time in their own, niche, specific way. To assume all engineers will become AI-pilled like you and hop on Claude Code whenever asked is the wrong approach. You can make the greatest tools, but if it doesn't fit in the behaviour and the way of working of the engineer, they will simply discard it.
I see this issue with clients and prospects all the time. A client's team of 5 produces 1 foobar widget in 2 weeks. That's 50 human days spent. Then, someone demonstrates the same thing can be produced (at an equal ot higher level of quality, mind you) with AI in 2 hours. Management might celebrate but teams will continue programming by hand as they always have, now asking ChatGPT about their build tool errors instead of using Stack Overflow.
Handing out tools and telling them it's cool is not enough. You'll need to understand, work with, and guide the engineering teams properly step by step. You'll have to change their behaviour. That does not happen overnight. Unfortunately the present approach is to drop the people who don't perform in the new era of AI-assisted software engineering. That is not the right approach in my eyes.
Can't think of a better poster child of complete corporate waste that benefits no one whose assets should be seized and redistributed to the masses.
For the amount that Meta wastes on LLM spending you can pay for things like universal childcare, public community college, and providing free lunch to all public students.
If you care about things like money, look up the dollar returns on feeding children during their development or when you tell families they don't have be an economic burden for simply existing.
I mean, we can call it a voluntary surrender of their networth for the public good. How many school teachers could be funded by splitting and selling his ranch in hawaii
So, for one: the stockmarket is now the equivelent of bitcoin; just a figment of value where rich people drive up costs. Just like a car is _invaluable_ to you not because of it's material value but because what it does out strips it's raw goods, facebook is mostly a bunch of tiny bubbles.
So you ask yourself, _if this thing disappeared tomorrow_, what would be the actual loss. It's definitely not it's valuation.
How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
It's very easy to say that someone/some oeganization's wealth should be confiscated, yet I have yet to see those proposing it actually putting any of their own money where their mouth is.
> How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
At least in the society I live all of those are partially paid by me through taxes.
I'm very glad to do it since the existence of kids school lunches, free healthcare (including for the terminally ill), and free universities make my life much better since society as a whole is better off. Even as an immigrant which did not use any of those services, I'm glad to do my part to pay for them, it's just the cost of a good society.
People like Zuck are incredible outliers when it comes to wealth and most people are basically underwater financially. A lot of us do send some money to the food bank or the local arts initiative but the amount left after the bills for living a regular life come due is a lot less for me than a guy with over $200b and no good ideas. That's a crazy order of magnitude difference, a thousand times a thousand times 200k which most people aren't even making.
For all of the pretty gross stuff it seems Bill Gates has been up to, at least he did spend a lot on stuff like mosquito nets and AIDS prevention.
> How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
A marginal rate of 52%. Well, technically we pay for the education rather than student loans here, because it's more cost effective without the middlemen (though we do get the occasional think tanks suggesting we change that).
I think that over long developers will desperately be needed to handle AI.
In my experience, within weeks now concepts written in stone get shattered and the next paradigm has to be used in order to max out AI in an development environment.
What is the case for AI? To handle basic work? Augment the work? Add work?
Why I think dev will be in a good spot if they adapt is the simple fact, that while laymen are using ChatGPT etc. every day, this is like driving a Tesla vs a formula 1 car.
If you take ChatGPT away from the laymen, they are helpless with IT. Devs aren't.
AI isn't static, and every turn evolves into complexity, only devs may handle when they adapt to frequent paradigm shifts and go into high level mode.
It will be again the interface between men and machine, laymen and AI. The gap won't close anytime as expected (The programming manager - remember 6 month ago?), but widens more and more.
What I see is that in day to day work many services have arms race with AI updates. The managers are more and more overwhelmed by the workload but how to automate systems is still devs' area to shine.
The business case is still hidden and unclear, but only one aspect is clear to me: low level programming is mostly configuration work now and bug fixing for AI very seldomly now.
If AI was a productivity boon: wouldn't a company employ the same or more employees to capture more market share at a competitive edge. Shedding employees because they are more efficient seems like shooting yourself in the foot to try and stand in the same spot in a race.
AI should have caused a job market boon: because less skilled employees would have been more hirable/useful. That this is not the case leads me to suspect that AI is an excuse to reduce employee count, but not the root cause.
This technology isn't even a decade old. It hasn't even been really useful until the last 3 years or so. Why would you expect it to be transformative already?
yup - its the j curve of productivity phenomenon in action. and we'll see smaller companies reaping benefits faster becasue they can start with AI native development.
I feel as thought Meta, compared to other tech giants, have a vested interest in saying that AI failed, as they are the only major tech company that has almost unequivocally lost the AI race.
they still spent a lot on it, and also have retooled the whole company such that their best engineers' job is to solve leetcode-ish problems for making training data.
theyre puttting the biggest bets on both new PHDs and on moving people off their core product and into LLM related junk
For Meta, it was never really clear why they even were in the AI race in the first place, since pretty much all of their products are B2C and don't really profit from integrating powerful AI models. And for all of their internal needs, they could easily use models created by someone else, which is orders of magnitude cheaper than trying to compete on building your own SOTA model.
Zuckerberg is proof you only need to get really lucky once. He's basically flailed at leading facebook ever since, going all in on monumentally dumb bets and erratically making headcount and project decisions, and still he somehow keeps getting richer.
Even if only for how well Instagram and WhatsApp were managed post acquisition, your message doesn’t hold up.
Most large tech companies buy late and struggle to integrate the products and teams they acquire without creating massive damage. He didn’t. Twice.
I am quite far from a fan of what he brought to the world (privacy nightmares), but to say that he’s a proof you only need to get lucky once is quite far off.
also: face tagging in photos, the fb app platform and the like button pioneered concepts that we take for granted today. also, how product is done internally, growth metrics, all borne from facebook.
The like button is an ingenious and insidious invention. It compelled every other external web property to willingly load fb sdk. It’s mind blowing how effective it is.
I am morally against Facebook, but anyone that says zucky was lucky, as a tech person, likely just sour grapes.
IMO the actual prescient call he made early on is to hang on to 60% of the total voting power of the company. I have to imagine there are several points where he'd have been pushed out quite a while ago if he hadn't have done that.
Also, it’s interesting to note that in both cases they where two very lean companies (Instagram had 13 employees and WhatsApp had 55) with pretty a pretty large and engaged user base (30 million for Insta and 450 million for WhatsApp)
Just because he fails publicly sometimes does not mean he only makes bad decisions. Facebook is still incredibly successful. The market may apply a Zuckerberg discount compared to other tech companies, but that is not the same thing as saying he has been a bad operator.
As a founder who has gone through scaling processes, I am quite impressed at Zuckerberg's ability to remain relatively competent and relevant at the top role throughout the whole journey. The skillset required changes quite a lot in each phase.
There are very few cases where the archetypical young garage-founder has stayed on top in the long-run.
Don't mistake tenure for competence. Meta's stock structure is such that Zuck cannot be forced out by the board or the shareholders regardless of how many mistakes he makes.
HN is so extreme sometimes. I remember when Llama was gaining popularity, everyone here was praising him for releasing open models. Same for other works in this area. Also remember praise for some VR devices.
And it's all back to Facebook, nothing in between. It doesn't look any different from general population influenced by TV and other social media.
> I remember when Llama was gaining popularity, everyone here was praising him for releasing open models. Same for other works in this area. Also remember praise for some VR devices
Neither of which have been great commercial successes. Open source LLM's and VR devices have have been cool products for HN readers, but haven't made much impact on the public or Meta's revenue.
Whilst Instagram and Facebook Ad revenue is coming in it means Zuckerberg could spend billions pivoting the company to country and western themed steak houses and the share holders wouldn't (couldn't) do anything about it.
Just a stab in the dark, but it's likely because FB and meta are wildly successful businesses, and it's almost impossible to attribute that to one "lucky" event at the inception of the business.
If I'm being honest I'm actually kind of shocked at the engagement this comment is getting and the fact it's still visible. Like him or not, if you're capable of being objective, he is doing the opposite of failing...
You know that a lot of acquisitions is equity hire. How do you get engineers to be excited about fixing something low level in Facebook, a hard task; you don't. Better to wrap those problems in green field projects and have engineers excited, and adopt that tech back into Facebook. Google done this forever (DeepMind). Seems like the work of a genius?
Zuckerbergs comment from 2010: «We have not once bought a company for the company. We buy companies to get excellent people.»
Getting a company to IPO is not about getting lucky once - it's making hundreds of good decisions daily. From hiring to product direction to fundraising.
Like or hate the guy you gotta give credit where it's due. (Same for Musk, etc.)
The man has its flaws but to deny that he has been a very good operator is just putting your head in the sand.
- acquired WhatsApp and Instagram and grew them into the most popular apps on the planet
- created an gigantic ad business that enables entire new types of SMBs to exist that rely on Meta to find their customers
- which is pretty much a money volcano
- bought a ton of GPUs right at the inflection of early LLM take-off
None of this was obvious when he started Facebook but obviously well executed in hindsight.
No one needs Meta AI. They have missed the boat with chatbots / agents / harnesses like Claude Code.
Can you picture anyone you know stopping using the AI services they currently use for an AI service provided by Meta?
They have such huge opportunities in selling access to their data centres to Anthropic etc, and improving their own ad models for better targeted adds (using their proprietary data and their own infra!), it is maddening to watch them try to make SOTA LLM models and harnesses no one will ever use.
CEO Mark Zuckerberg recently dispatched a small team at his company to create a smartphone app similar to Polymarket and Kalshi, the New York Times reported on Tuesday, citing two employees with knowledge of the matter.
The app will probably rely on a video game-like points system instead of users wagering money, though the company has not ruled out betting real money eventually, according to the report.
Meta will continue trying to build a platform that they can control. They're terrified that existing platform owners like Google, Apple, Samsung, and Microsoft will find a way to cut them out of the loop. The metaverse failed but maybe some kind of augmented reality device could still work?
They're pushing hard for a unified platform. For WhatsApps new username feature one can only choose from usernames not already used on instagram or facebook.
They're going to try to push those stupid glasses next. Not convinced they'll succeed since it's not just the technology problem they need to solve. How do I find PMF for a product which encourages others to physically assault my users, while also not having it banned in various countries? We're talking about a societal shift that would need to happen and nobody's going to trust Meta again where that stuff's concerned.
I don't want to be a Zuckerberg hater, but is there any thing Zuckerberg has done or said in the last few years that wasn't mundane, reactionary or self-aggrandizing?
In other words, was there a single decision or take he made that turned out in his favor?
> said a review of a recent data security incident with the company's controversial mouse-tracking software indicated that no employee data was included in AI training.
That's... not quite right. The employee data is used in AI training and is intended to be used this way. But despite not correctly ACLing the data for a couple weeks, it is believed it was not accessed inappropriately.
So what happened to Meta after those successful llama 3 model releases? They really made competing models back then. If felt like they have right people, strategy and good results. Now it feels they have neither of those…
You'd have thought that Zuck's previous failures to make the things he dreams about (e.g. Metaverse, decent in-house AI) materialize might have made him a bit more cautious about betting the farm on things that don't exist, especially when he's expecting someone else (the AI agent folks) to make it happen!
I suppose you have to admire the conviction: I'll fire my developers today because REAL SOON NOW I'll be able to replace them with AGI!
If a Meta employee screws up a major project, what happens? What will happen to the executives behind these mass firings and realignment - executives of one of the very top SV companies whose job is dealing with the landscape of disruptive technology development and overreacted to the latest thing? What is the standard for them?
I can't imagine they'd ever get enough people to opt-in to mouse tracking to generate enough data for their model training. It's either a DOA AI project, employees will be pressured into accepting it (making it part of their review), or their managers will accept the monitoring on their behalf.
with coding, you have sort of a framework for doing it right, if you have good specs, good testing practices, strict grounding in expected results deterministically, good linting, etc... this is much easier to automate with AI for the coding part within that assuming you did your homework around it... i don't have experience with all the business layers but it seems a bit more nuanced and fuzzy as you get away from that "harness" of sorts as it doesn't have to work in the same way as code for execution and evaluation... and even if code works, it still needs tastemakers in the final ok. maybe the taste maker ability still needs a lot of work/scale to be feasible, idk, like its still earlier than later on that. maybe Elon already cracked this to an extent given his automation in various companies.
it is amazing that meta continues to burn money and keep afloat. ads DO really make money, especially when they are going to overtake google in this [1].
to all the critics, i would suggest letting him/them cook. the snowden-era privacy concerns was exactly around ai getting trained or using personal data, and they have a treasure trove of it.
their money dump do sometimes also spend money on r&d unrelated to the current mindhive topics. we never know what the scatterbrain approach may give birth to tomorrow.
do be critical and vocal on how they squeeze money out of other areas or how they increase their revenue by ruining our lives instead.
Reminder: the "expected" result is replaced all programmers and white collar employees. Even they don't necessarily state it explicitly, that's what is in their minds.
By "AI agent development" does he mean using agents and such for in-house coding? Or does he mean their initiative to train their own agentic model based on the data he's having a bunch of employees generate?
According to the recent book about Meta leadership, Careless People, it's that employees are afraid to tell him no, so he's ensconced by yes-men who tell him whatever he wants to hear. He probably has no grasp of market and product development realities.
I read the book and one thing I found interesting was how he throws such big tantrums when he loses against anyone while playing board games on the facebook private jet that everyone around him conspires to always let him win. Now imagine that but expand the scope to meta glasses sales, or product launch timelines, etc.
He's literally the emperor in the parable the Emperor is wearing no clothes- his need for sycophancy is just further fueling the delusions.
I never noticed that the Emperor was a vain man who couldn't admit his lack of wisdom (because only wise people could "see" the beautiful cloth, and he pretended to be able to see it), and that he was surrounded by yes-men (of course he was surrounded by similarly vain men who had to pretend to be wise to keep their positions, but I didn't notice how this turned them to yes-men).
Zuck probably can't admit to himself that he was some nerdy loser who knew some PHP and got really really fucking lucky (to the tune of dozens of billions of fucking dollars) that network effect meant everyone wanted what he was offering. I'm guessing he thinks those billions must be proof that he's smart... So smart that he's unbeatable at any board game.
> how he throws such big tantrums when he loses against anyone while playing board games on the facebook private jet that everyone around him conspires to always let him win
It's hard to believe that that is a real person and not a fictional person being written against some trope.
Facebook is already renting out their excess GPU capacity. Nvidia is giving out GPU capacity for investment stakes in startups. OpenAI is trying to get the government to take a 5% stake and needs to make the same amount of revenue within 3 years as Google does now or else they're going to default on their payments to Oracle. It may not be the same as the 'dark fiber' from the Dot-com bubble but it sure seems similar. Especially when this hardware is purchased ahead of delivery and ages so quickly it's not even clear it will be reusable after a few years. It's the scale of the investment in this stuff that is the issue. In 3 years, AI spending is off the charts compared to the railway bubble, Dot-com bubble, Roaring 20s, or really any other bubble. I don't think it's going to end well.
Or: you wasted too much money on failing to replace yourselves so now I have to lay you off. Which is one of the two possible grand outcomes of the AI bubble, which both result in laying people off, because that is all these companies know how to do as a response to stress.
I am not sure that it has to be so zero sum. The AI truth is probably somewhere in the middle; it probably doesn't replace software engineers and it probably won't be deleted as completely useless. My current feeling is that it's a powerful tool I'm happy to pay to use; it doesn't replace me, but it makes it easier to do higher quality work. It feels a lot like IntelliSense, or faster compilers, or getting a 32" monitor. That probably doesn't sustain the bubble, but it's something that people are going to be poking at and making money off of for a long time.
I agree that people are investing as though the world is going to run itself while the ultra-wealthy run off in yachts to compare sizes. If it wasn't AI, it would just be tulips or something. That's just how people are. But maybe they'll be right, who knows.
there are plenty of orgs where writing more code is not a good thing, in fact it's the last thing they want. but yet these orgs would still employ 80%+ of all developers, not necessarily to write code though.
Investor money being pissed up against the wall, for little purpose? What else is new.
I'm amazed the backers are on board. I wouldn't be. I'd rather invest in dead tech like coal, and soak every last drop of profit, than back this kind of bogus shit which is proving both more expensive, and less workable. Well, maybe not coal, but for goodnes sake, can we cut to the chase where AI investment tanks and move on with our lives?
I'm sorry if it's a non sequitur but I feel even beyond superintelligence/AI/LLM whatever of the last few years... they've always done this, it's always been somewhat hamfisted
Examples abound of "I reported Nazi hate page. Didn't violate community guidelines. I called my friend a jerk, jokingly, got a month ban
For years. Not restricted to when ChatGPT et al arrived on the scene
(Because, AI in theory makes sense. If you want to monitor things at scale you might use AI - however that's defined - to make your workload easier. When is an account being hijacked? When are bad actors infiltrating the system? Or whatever)
I think there are seriously misplaced expectations here. The primary role of AI is transference of effort, while "increased productivity" is just a side-effect (since computers are so much faster than humans at highly repetitive tasks). It's about not having to directly do X anymore (or as often), even though it may take a few rounds to get X to a satisfactory point. But even if following up is needed, most of the effort budget can then be used for Y.
Also those with very heavy investment in AI are looking for bonkers results, which is the cause of their disappointment. They need to reduce their expectations. I for one am loving the results so far.
Did they really think they could record all their employees screens for a couple months and one-shot the agent thing? This is like junior engineer "let's refactor this monolith" levels of delusion.
My instinct (for better or worse) is usually contrarian. Most people seem very skeptical of what Meta is doing with AI. But, what if, in a way at least, it makes sense?
Maybe Wang has correctly identified that the programming and agentic ability that Anthropic and OpenAI models have has largely come from armies of software engineers creating massive datasets by writing out coding and agentic problems and solutions?
So he told Zuckerberg that. The reason it may be turning into so much friction is that at companies like Anthropic or OpenAI, training engineers were either hired specifically for that purpose or probably mostly handled through contracts with third parties (which again, hired them to train AI). And honestly many of them may be overseas or just happy to have a job in a difficult period. But anyway they wouldn't have very high salary expectations etc.
But Zuckerberg already had 25000 engineers. Why not take say 1/5 of them and get them working on the the dataset? The problem is that those engineers were hired for different prestigious highly paid positions at Meta/Facebook. They were not hired to do tedious grading of AI answers or quiz construction.
But Zuckerberg either has to do this, or spend additional billions on doing it all with external contractors. A third option would be to try to create a massive distillation operation. Or just hope that his engineers could invent some magical new training trick that manifested the agentic and programming skills without the large scale human input.
Or he could release a model trained largely by existing open weights models. Which without some huge breakthrough probably has no chance of surpassing them, so is pointless.
I think most of the substantive criticism of Zuckerberg has been about burning funds. If he gives up the "your job is to grade AI homework now" plan because his engineers refuse, he would need to go through third parties. The additional billions and billions this would cost would create more pressure on the bottom line and shareholder pressure.
It would also give up any potential advantage that Wang may have optimistically sold the operation as, on that using "real" engineers as opposed to lower paid data labelling engineers might result in a higher quality dataset.
At some point, model architectures that don't need such massive datasets or can be created automatically in a way that advances the frontier will probably come about. But right now it doesn't exist.
Further, the way AI works currently, business advantage from AI comes from encoding existing internal intelligence and knowledge. Meta's massive engineering corp effectively has that in their heads. Having them create these datasets is possibly the only way to leverage this knowledge asset in this paradigm.
I guess the problem is it means forcing thousands of people to do a different job from the one they were hired for.
What's the end goal? Meta-specific engineering, with baked-in knowledge of how FB, Threads, and WhatsApp work? General and/or coding products to compete with Anthropic and OpenAI? Some special Magic Thing which only Meta can invent which will bedazzle Meta's users?
You don't need giant datasets unless you know what you're going to do with them. OpAI and Anthropic are having enough issues making their products profitable. And those are, if not beloved, then at least respected, with a real, if patchy, reputation for usefulness.
What was Meta's pitch in this market? There were hints of interest when LeCun was still doing original R&D, and there was some distant possibility of a next-gen revolutionary product.
But now the goal seems to be to flail around doing something incoherently AI-branded with no obvious strategy.
The troops are being marched around, but no one knows where the battle is supposed to be.
They are making more money than ever before. Maybe Meta leadership doesn’t really care about having a coherent strategy at this point. They can afford to flail around to see if something sticks. Reminds me of Rich kids who have ability to travel the world and find themselves before settling into a career
One problem is that the AI agent market is fiercely competitive. Why build when you can buy? For the foreseeable future there will be a number of competitive models on the "efficient frontier" and I don't think one vendor will pull ahead.
In that market you can build a model and spend a lot of money on it and at best get something that's on the same frontier as everybody else but just as likely end up with uncompetitive models like the ones they have now.
You might save a bit running your own models, doing your own inference, etc. Why not take advantage of "last mover advantage" and buy whatever is best when you need it and figure the odds are good that everybody else is going to buy more GPUs than they need and as a large customer you'll be able to buy in bulk at fire sale prices?
That makes sense in a way, but remember that Meta had previously seen some brief developer glory in the initial Llama release. Going the off-the-shelf route would essentially be giving up on being on the technology frontier in this area, and not monetizing their knowledge assets.
>I think most of the substantive criticism of Zuckerberg has been about burning funds.
I'm not in the org myself I know some Meta SWEs tangentially. My understanding is that the biggest criticism is just the chaos of it all. Jumping constantly from one thing to another like headless chickens and accomplishing nothing.
It created an environment where it's kind of impossible to plan and progress your career.
While I mostly agree with your post, I do want to point out one thing:
> Or he could release a model trained largely by existing open weights models. Which without some huge breakthrough probably has no chance of surpassing them, so is pointless.
This seems to be categorically untrue. Composer 2.5 is a substantial improvement on its underlying Kimi base model.
If that is backed up by benchmarks then maybe they should imitate whatever Cursor did. What did they do?
They may eventually have to do that. Or they might be starting with an existing Llama model. Maybe I should have said "huge breakthrough or additional dataset".
We're getting closer to the bubble-bursting peak. Yet another tech corp that fired a ton of experienced people hoping to squeeze more $ for a quarter, finding out AI hype is not all it's cracked up to be. Loving that for them.
Second thought, I wonder if zuckburger has turned into the Cramer of tech. First, the metaverse fiasco, now another massive overinvestment into a dead-ish end?
Ahh the classic bootlicker's bias, this is how oppressed developers move and click their sad cursor hoping to keep their job while they train their AI replacement!
>"For people who are comfortable, that's great, they can contribute to this kind of great human survey. To people who are not, it is not an issue,"
That's a great question. Some form of union has already started in some companies, as far as I know. But not many employees have joined those unions. Probably because most employees have high income and don't really feel like they need collective bargaining power, compared to other low-income laborers.
The modern trend is to think intelligence is generative “like compression” or “predicting next in sequence” rather than iteratively reducing uncertainty, like those fault tolerant humans.
Compression can be defined as reducing uncertainty. If you can predict the next sequence you can compress it to 0 bytes using arithmetic coding. Reliable prediction is what enables compression and it's the link between compression and AI that everyone is talking about.
No one ever in comp sci says artificial intelligence is "like compression", they correctly state that "artificial intelligence IS compression". It's absolutely known and accepted that artificial intelligence (defined as predicting outcomes with a measure of certainty and taking chosen actions towards goals using those predictions) has equivalence to compression in a very hard science way. The hardest part of artificial intelligence is compression and the remaining part, the choice of actions based on predictions is just a tree search to a goal.
Although there is a difference in accrual earnings and cash flows. I was wrong with the number provided which although not as big as 1.5B of AI spend, is still a comparatively large number itself.
I have written an more in-depth comment if that interests ya (in a good faith discussion and please be kind to everybody)
also please don't call @kentonv an idiot and please read the HN guidelines[0]: Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.
Mark is really a bad leader with a mwah mwah vision. He is maybe correct in some things. But the execution is really really poor. Plus he does not have followers and believers. He only got money that can simulate followers to a certain extent
If it was still possible to get verbatim results from Google then I believe "mwah mwah vision" would have been an authentic Googlewhack pointing at this comment thread.
Actually, agents have been developing incredibly fast. It is just that Zuckerberg has some unrealistic expectations; the man is a lunatic.
Over the past six months or so, OpenAI's internal team has completely shifted from being heavy ChatGPT users to using Codex. Once you start using an agent like Codex, it is very hard to go back.This shift is truly transformative.
I am also aware that some of the consumer agent products on the market are growing very rapidly, such as Manus and GenSpark. Not to mention Claude Code and Codex.
I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened. Do I write all my code with an agent now? Yes. Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not. I can produce more code than I used to, but if I want it to be good, to be stable, to do what the product manager and designers want, it's only about 2 to 3 times more code than before. And that productivity is impacted by the fact that I'm reviewing 2 to 3 times more code than before (and you have to review, even more so now than before, because if you just let opus or gpt 5 do its thing, you'll get some terrible results, and I've found a lot of engineers on my team are just letting it do it's thing without a lot of iteration).
The thing is: I could produce 2-3 times as much code as before _without_ an LLM, if I didn't care about my colleagues' ability to review my output properly.
Lines of code are a liability, not an asset. You want as few of them as you can get away with, without compromising the actual asset: the functionality.
A huge part of the job of Software Engineering is producing the right amount of code at the right time.
> Lines of code are a liability, not an asset. You want as few of them as you can get away with, without compromising the actual asset: the functionality.
> A huge part of the job of Software Engineering is producing the right amount of code at the right time.
Absolutely true, however my experience says that the correlation between "good software engineering practices" and "positive business outcomes" is, at best, small.
120 kloc mostly from one single developer copy-pasting and keeping non-compilable code for an obsolete target "for reference" for a decade, becoming both a ball of mud and a whole pantheon of god classes? No unit tests, no code review? Won awards.
Properly engineered, mandatory code review, mandatory unit tests, dev meetings to knowledge-share? People with the money said too slow, closed it down.
(Sometimes people bring up how bad Musk's code was at PayPal. I never bothered investigating. Successful product though, wasn't it?)
16 replies →
> Lines of code are a liability, not an asset.
I have been saying this for years, I once had a heated argument about a small system of maybe 1000 lines of code that was technically superior and more scalable but was freaking 1000 lines of code to maintain compared to the quick and dirty 10 lines of code it was suppose to abstract and make generic (for future use of course).
That with also countless debates over insignificant features in frontend apps at the cost of extra code. Frontend code is very susceptible to this maintenance cost dilema.
Many developers are too focused on delivery value compared to maintenance cost. It is unfortunate that non-technical management can see value delivered, but not maintenance cost incurred. With LLM-assisted code this has become many times worse.
11 replies →
This is the thing. You "spend" lines of code, you dont produce it. The produced part is the outcome - a functional feature, stability improvements, some business outcome. Measuring productivity with LoCs is like measuring output with cash burn.
1 reply →
I think that coding reviews are no longer feasible as they used to.
The pace and expectations have increased and a human can barely cope with reviewing its own code, let alone colleagues'.
They are not going to disappear in critical aspects of a codebase, nor shouldn't, but the industry will eventually reward self sufficient individuals able to keep the pace, harness and run adversarial reviews against the design and implementation autonomously.
I'll also say the harsh truth. A well implemented adversarial flow will do either better than your peers or will deliver 95% of the value at a fraction of the cost.
The industry has never valued product, let alone code quality except in places they are core to the business.
Otherwise you would not have MIT-bred leetcode ninjas writing react/tailwind bugged monstrosities at half a million/year for billion dollar products.
5 replies →
"A huge part of the job of Software Engineering is producing the right amount of code at the right time."
But the whole job of the owners is to lower the costs as much as possible to produce the product, especially in an environment with higher costs to borrow (higher interest rates at the fed)
> A huge part of the job of Software Engineering is producing the right amount of code at the right time.
I'd go further and say that usually the goal is to use as little code as possible without sacrificing readability.
Brevity is compression, and compression surfaces the salient points of a problem.
Elegance often comes down to brevity.
This is one of the things I ran into early on. LLM needs to compute the determinant of a matrix? Sure, just spit out some huge hyper optimized implementation of it. Good luck maintaining that. Slapping "use industry standard open source libraries for common functions" has improved the quality of LLM output for me by such a large margin.
So it's like jazz... it's the code you didn't commit that matters?
>I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened.
I find this somewhat puzzling. I thought things were moving quickly, but at this time last year I couldn't even get Claude (using Cursor) to spin me up a service skeleton that would compile, let alone do anything meaningful.
I know it feels like a long time somehow, but it was only between November and February that things started to actually somewhat work without significant hand holding. Even now, it seems like we're still figuring out how to fully leverage the current models and tooling, even in organizations that have largely gotten on board.
It's not all that surprising that people were worried and believed this. The AI companies and infrastructure companies partnering with them have spent a lot of money and time trying to convince people this is the case year after year. The critical clue people miss is that everyone claiming that has very clear financial incentives to convince people that's the case even when they know it isn't. Anyone who was actually building with LLMs and judging for themselves based on its performance knew fully well that wasn't the case year after year.
10 replies →
coding harnesses improvements mattered more than llm improvements this past year. You could solve problems on claude sonnet on claude code that you couldn't solve a year prior.
1 reply →
> at this time last year I couldn't even get Claude (using Cursor) to spin me up a service skeleton that would compile, let alone do anything meaningful
I've been using it to do this for 2 years now. And many people with me. The change you mention is one of is primarily one of Overton windows, of vibes.
7 replies →
Do I write all my code with an agent now? Yes. Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not.
I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them.
Some devs just 'get it' and thrive, leading a team really well and building a great culture. But a lot of them don't, especially if they don't get the support necessary to understand what changes when you move from IC to manager. If the team (or agent swarm) isn't performing well it often isn't a problem with them. It's a problem with the new manager still trying to stay on top of everything and micromanaging all the things. Alternatively, the new manager is completely hands off and only appears at a check-in point (one-to-one, agent completes a task, etc) where they crap on the work and get cross.
I have no evidence for this, but I'd guess that putting developers through some sort of management training would make them much better at using agentic swarms.
Absurd.
We are beset on all sides with companies declaring agentic coding a failure and here you are stating as a matter of fact some teams “thrive” with this probabilistic expensive approach to approximating working code?
All the while concluding with “I have no evidence for any of this”.
45 replies →
> Some devs just 'get it' and thrive, leading a team really well and building a great culture.
I don't think it's this because the outcome you get from AI isn't controllable. You can give it the best prompts and design suggestions and it'll still give you completely wrong or horribly written code.
If you were a manager and one of your reports kept producing completely wrong and horribly written code that other folks on the team keep bringing up as problematic in PR reviews or privately, that developer would eventually be fired for someone better.
But in the AI case, there is no replacement because all of the LLMs have severe problems.
4 replies →
> If the team (or agent swarm) isn't performing well it often isn't a problem with them. It's a problem with the new manager still trying to stay on top of everything and micromanaging all the things.
You see problems in the results? Simple, don’t check the results!
1 reply →
Yes, an agent does feel like a very eager and very knowledgeable but also very clueless junior developer sometimes. But no, working with agents is not like leading a team of developers. You don't have to make sure an agent stays motivated, give it the feeling that its work is valued (I can let an agent spend some time building a prototype and then decide not to use it, if I did the same thing with something a junior developer took a week to build, that probably wouldn't be a good idea) etc.
> I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them.
I disagree.
From ICs that I lead, I expect that they learn fast. On the other hand, LLMs are basically incapable of improving.
Another issue that I can see is that I don't particularly like eager new colleagues who come up with (hallucinate) wrong answers. At the beginning, if you are uncertain, learn, and if you have questions that learning does not answer by itself, ask questions. But strongly avoid hallucinating answers. New colleagues can be taught that, LLMs not so.
3 replies →
> Some devs just 'get it' and thrive, leading a team really well and building a great culture.
I think what makes a dev well suited for AI isn't the same as managing a team. What really helped me get productive is having to write a lot of user stories and acceptance criteria with the wisdom of being a dev tasked with implementing them. Also, being on the refinement calls, answering questions, and updating requirements/AC is good feedback for authoring better requirements. If you're good at authoring requirements, checking output, and communicating corrections concisely then you can get the LLMs to sing.
I get why you think this but you are incorrect. Why? Because managing a team of people is very very different from managing outsourced contractors, and LLMs are much more like the latter than the former.
"I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them."
I doubt that. Management is mostly dealing with people, the actual "management"* part is not where developers moving into management roles typically fail, it's the people part. With agents you have the management part without the people part.
I think you are confusing manager with ICs. Managers don't really read or review the codes. What you are describing is where agents are doing all the coding and reviewing without people in the loop. I don't think the op is working with the code as black box. He is more about describing the higher IC workflow.
Whether we still need people in the coding loop is not a trivial difference
2 replies →
I agree there is a skill change but agents aren't people so managing them is very different. E.g. you can spin up 1000 agents. Get them inti tight loops and get them more context and so on. A manager doesn't do that with people.
1 reply →
Completely agree.
The job isn't to write code. The job isn't to architect. Those are means to an end.
Anecdotally: it seems like the most principled developers are having the most trouble adapting to agentic workflows.
In my experience, it's the opposite. People who become dependent on AI agents are avoiding human contact, and people who become managers are seeking it out (for better or worse).
I disagree with your 2nd assertion. Even engineers who are less tech lead style engineers, can gain a significant boost in productivity by being able to quickly run through POCs and build an understanding of surrounding areas of their work, so they are able to contribute more.
For eg I am able to make React changes much faster and the changes are higher quality, given frontend dev has never been my job role. I’m able to spin up test harnesses, write throw away glue code, test against large datasets, etc
2 replies →
> make them much better at using agentic swarms.
Do you mostly use agentic swarms? If so, I’d be curious of your use cases. People talk about “managing agentic swarms” a decent bit, especially on LinkedIn. I just don’t see how they are the best solution for majority of development use cases. At best they seem like using only a hammer to make sculpture.. or a sandwich.
The problem often is that the new manager can teach their hires to write better code and train them to be new managers. Given that a team leader in such a scenario is often checking for correctness, I am not sure if current LLM based AI will ever be able to do it. We need new AI for it.
There's just no escaping from making decisions. Agents can't reliably make them as many rely on tacit knowledge which isn't written down.
This bottleneck will always have to exist unless companies just accept AI defaults, with predictable outcomes.
I find it interesting that when drawing this parallel you mention that some devs 'get it' and 'build a great culture'; I think this is exactly where the analogy breaks down. Good managers get great results from people (and for people! they are linked).
Good AI managers are just running optimization loops at more declarative levels. Yeah, you need to get comfortable with less personal review of code for both, but I think the differences outweigh the commonalities - it's much easier for someone with a more 'traditional' IC model to be successful with agents then they would be with management, and I think most (good) management training would be entirely irrelevant. Parallels are maybe tighter to higher IC progressions.
Nothing you said makes any sense.
LLMs are driven by the text you enter into them and they are never fully autonomous unlike an actual employee.
You cannot train an AI and then at some point just let it loose.
[dead]
> Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not.
Ignoring instructions - whether in AGENTS.md or my prompt - is the worst of it, and it routinely happens. It just waives things that I explicitly told it to do as part of the design.
Vibe coders (in the true sense, zero oversight) claim that you just need to prompt it carefully. That's completely untrue when faced with your careful prompt being ignored.
I even have "don't overrule me without asking" in my global AGENTS.md, and it simply doesn't do that.
These are word generators, not agents, I’m really not sure why people think they could be capable agents (ie independent) when they consistently ignore instructions, generate the wrong things and then double down when questioned, etc etc.
You’ve been sold something that simply doesn’t work for the purported use case (intelligence) and instead is like a stupid database of all world knowledge with the appearance of intelligence.
Useful tools at times (if you bear in mind their limitations), but not close to intelligent, independent agents.
1 reply →
Your context isn’t to give it orders, they just don’t work like that. Your context (AGENTS.me, skills, per-request context we are sending in for each request to bots) is to give it the info it needs in the language category it’s trained for the answers you want; you have to give it a clear instruction each prompt. Basically, when you have a long session, you can see this by saying, ok, now moving onto another thing, blah blah blah (implicitly ignoring all previous instructions). It can even back fire - nagging too much about don’t skip tests in the context can make it slip into the linguistic space where there is some emergency and faking the results might be justified (I imagine there is a certain amount of training out there “just making the tests pass for now, will fix later, I promise.” If you rarely mention tests except “this one is failing, please investigate what is going on” (an informational outcome not a test outcome), it doesn’t really “cheat” (tho it can leap to conclusions as always). The tests need to be some deterministic step in the process anyways, tests don’t need fuzzy word directed search capabilities. But the models just don’t have the structure to allow feeding in a ten page set of rules and follow them. You can add a step to say, please check this git commit for compliance with the 23 rules in this standards file, and it will work better to catch the gaps.
3 replies →
I’m convinced the magic bullet is deterministic checks. Linters, static analyzers, etc. Whatever you can do to create deterministic gates that the LLM simply must overcome to reach a “done” state, do it. Has been making a huge difference for my team, but sister teams are so invested in writing the perfect Make No Mistakes prompt that they just can’t see it.
Basically I treat it like a junior dev. We don’t get junior devs to write code correctly by cajoling them just right, we add CI gates. It still works.
14 replies →
> that I explicitly told it
Try writing it in first person instead of second person or neutral.
A while ago someone had a similar complaint on here and shared some example lines, and that popped out at me immediately. However much structure we've wrapped these in, they're still text generators trained on all sorts of things, and if you think about a narrative where first and second person speech would be used, try to imagine context: In first person, it's most likely a description of something as it happens or someone planning what they will do. But in second person, especially command form, you open up to the possibility of commands being ignored, misunderstood, or actively rebelled against.
Whoever that was back then did some quick tests and found the pattern held, first person got it to follow far more reliably.
2 replies →
> I even have "don't overrule me without asking" in my global AGENTS.md, and it simply doesn't do that.
You really need to look into hooks based on your coding agent. This is very much a solved problem as I demonstrate with
https://github.com/gitsense/pi-brains
I have a test repo
https://github.com/gitsense/gsc-rules-demos
that shows how you can block and warn and do other things.
You obviously can't have a "Don't make a mistake" rule though.
1 reply →
People will need to realize that these behavior are actually truly human in nature.
People thought they'd get their own persona through AI, they get a sum of the best and worse of everyone.
Also noticed this. Their intelligence is very jagged. I’ve had them produce some highly optimized code yet fail to follow basic code guidelines.
It can't follow orders because it isn't thinking - even in a best case scenario, all you're doing with AGENTS.md is altering the probability distribution of outcomes away from things you don't want towards things you do want.
However that still means there's always some probability it will do things you told it not to, it's just reduced
In my limited testing Fable is far better at obeying CLAUDE.MD than Opus is.
6 replies →
The short leash method is the way to avoid this.
Companies are putting a ton of effort into getting to that point of having agents do the work unsupervised. Whoever gets there first is going to be the winner.
I personally don't think it's possible and I haven't written a line of code since Sept 2025.
There's an AI psychosis going on right now, especially among the execs or management class, and we all gotta nod our heads in agreement and burn through tokens.
Agents already run unsupervised, and they can code unsupervised too. The real question is what worthwhile work we should point these capabilities at. Nobody has really cracked that yet.
16 replies →
If you have runway, it’s a good time to start your own thing or join someone who is. Personally, I cannot force myself to wade through slop PRs from careless coworkers. If that’s the job now, I’d rather run a hotdog stand or something.
Luckily, I don’t think things are that dire. I think the companies issuing AI mandates are manufacturing sawdust, and even if it works, it would just enable them to burn through customer goodwill in record time as they make user-hostile decisions free from engineer pushback.
These are going to be a few tough years, but I think the opportunities to start something new are everywhere.
4 replies →
I don't think it's possible either. DHH pointed out in the most recent episode of Rework that AI removes a lot of the barriers to shipping code, therefore making it possible to build in lots of different directions in ways that was prohibitive to many organizations in the past. But this isn't necessarily a good thing, companies still need to understand what to build in order to ship a cohesive product. AI is great for prototyping and refining use cases in ways that are far superior to static figma designs, etc., but it is not a replacement for taste and execution.
But a slop machine that haphazardly shoots features against the wall to see what sticks still isn't a winning product strategy in 2026. And the problem I see increasingly is that so much energy is being focused on how to deliver with AI internally and externally that is not being expended to advance a company's product. I believe more and more in the idea that for many startups and companies, the actual "customers" are the investors and the product-market fit that companies seek is the product of the company itself, because this is all being driven from the top down, not by customers and users in the market asking for AI features.
5 replies →
I'm really struggling to get an agent to write code I'm happy with. It's mostly pretty awful.
I've a fairly simple c# coding style. But simple is proving a bit more difficult to convey than I thought.
I get it to produce code. I then have to spend along time convincing myself it's correct. If I don't I end up embarrassing myself when a coworker reviews it, questions it and it's obvious I don't properly understand it.
This is really starting to screw with me mentally. It's like everyone in the world is saying they can fly by flapping their arms (dark factories). When I try I just stay in the same spot burning a lot of energy.
I don't think everyone in the world is saying they can fly by flapping their arms. It's a small number of very vocal, very online, AI enthusiasts, many who have a financial stake in AI winning.
work from tests to implementation. Validate the tests, and work with the agent to ensure there are not more cases that need testing. Then you can let the agent implement the code and you can refine it until it's simple enough but covers the test cases. TDD is the only way to use agents effectively, imo
1 reply →
Measuring software performance by lines of code is like measuring aircraft performance by weight.
I have no idea why everyone seems to have forgotten this simple fact over the last four years.
Measuring aircraft production per weight doesnt sound like that bad of a proxy. If I hear that Boeing produced 300 kilotons worth of aircrafts more this year, I'd be right to suspect they've ramped up production.
2 replies →
"companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened. "
It never was going to happen.
Always the same story: https://en.wikipedia.org/wiki/Gartner_hype_cycle#/media/File...
> I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened.
Amara’s law: We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.
This continues to be applied to AI where people think is going to be next 12 to 18 months. Changes are coming but certainly not at the rate Zuckerberg and most people are thinking.
I believe this in general but the issue with AI specifically is historically in every cycle we've overestimated the potential of AI in both the short and the long run - people in the 60s were genuinely quite convinced we were close to AGI for a few years, as ridiculous as that now sounds.
This is why you get "AI winters" but we've never had a "steam engine winter" or a "railway winter" or a "petrochemicals winter"
3 replies →
Back in the day there was a mantra: best amount od code is 0. Now we have agents spitting lines after lines.
I am not afraid of my future. Even if one person can do work of 5, the amount of generated code will grow exponentially. And not everything can he vibecoded with 0 knowledge. There is a complexity that need understanding to change and optimize. For now :)
> Back in the day there was a mantra: best amount od code is 0.
It was true for that time, because producing and maintaining the code was done by humans with limited speed of comprehension.
Today, we might challenge this assumption (not saying its wrong or right), because migrations can be done in 1-2 weeks with hundreds of agents.
I think it doesn't prove much that it hasn't happened yet. Companies might just be moving slower than you think, and are still planning on doing it. And, in many corners, "don't manually write code" is being joined by "don't manually read code" as an attractive principle.
> in many corners, "don't manually write code" is being joined by "don't manually read code" as an attractive principle.
I'm pretty sure I know where the failure case on that one is. The reason we're still manually reading code is to catch the failures and edge cases that the LLM fails to; not reading the code doesn't magically make the code good.
The Yale economist Pascual Restrepo, who is well regarded researcher in automation, doesn't think it will happen for most jobs.
https://fortune.com/2026/04/04/ai-jobs-future-not-important-...
Sure but there's also little reason to think we'll be able to replace xyz role entirely with software next month, or year, or decade. It's easy to disprove claims about the future; it's quite difficult to make believable ones.
you might have to think the way through though and these companies are already being caught up with the huge token costs at the same time.
There was an interesting comment during the cloudflare layoffs (partially driven by the fact that the company was bleeding money also because of its token costs from one estimate being 5* million$ per month (I feel so silly that I accidentally had written/meant 500 and had kentonv do the stats on that part :-( Sorry kentonv!), don't quote me on that though)
The part was that there is only an enough marketshare in the first place. Cloudflare was doing some crazy experiments like operating matrix on cf workers and wordpress alternative and fediverse and so much stuff.
So they basically spent 10x the amount of token (and the token costs) and I imagine as such the reading code of that part was getting sidelined as the attractive principle you are talking about.
Yet the market can't bring an actual demand 10x times though. These are things which nudge a user slightly but the actual impact on user growth isn't 10x or even justifiable within some cases given the costs.
Yet at the same time driving up the people who actually know their stuff and firing them because of the token costs. The people who have actually mitigated some of the largest DDOS attacks and are the backbone behind cf cash-cow (enterprise payments) is the fact that they have had the experience and entreprise knowledge about these things, yet they are literally removing that by firing workers and oh replacing them with interns. (They got 1111 interns and fired 1100 employees or something iirc)
It's weird and I have talked to some people about it but there is a disconnect between what management is hearing about AI and the ground reality of things. Reviewing code is becoming the bottleneck but if you don't review code and are shipping things to production, then you can get fired as I have talked about in some of my other comments sharing a story about how a guy shipped code to prod and the response was "but claude generated it" and got fired because the company basically said, look we basically don't care if it was generated by claude but the responsibility was on you to check it (review) and because the commit was done by you, you are gonna be treated responsible and he got fired from his job.
Yet this was the same company which was asking its employee to play around with claude at their free time, the manager of the employee I talked to being the most automatable person, the company employees working till 1 AM because they were saying to management that things were fine but they were being burried under the technical debt,that employee that I talked to got honest with the management and told reality and the management treated them as a person who didn't know AI or were the odd one out.
Sooo I don't know actually to be honest.
TLDR: reviewing code is being treated as the bottleneck but it is also the only thing stopping your company from imploding under technical debt, actual debt because of token costs etc. I remain skeptical if we should treat it as a bottleneck or as a safeguard mechanism. After all, if nobody's in the loop then whose responsible?
Reviewing code isn't a bottleneck so much so its a safeguard mechanism in my opinion. Also things differ in corporate land and hobby land and I would prefer corporate to not be using the practices that I do with how I do things for fun in my hobby time.
Side note: Even more so, I think I am a LiteLLM security working group maintainer and I have seen first hand on how much damage it can do in supply chain even when things were done right from LiteLLM side and the fault was within the side of ironically a security product that they used called Trivy.
There are things which you can do to be better prone to supply chain attacks in general but there is no full bullet proof way of doing so and in such.
Caution (should) be taken when dealing with corporate systems and as such I sweat a little when anyone suggests code review to be completely eliminated. Things (are/can be) different in hobby/prototyping world though.
2 replies →
I have experienced and feel very much the same, and it is refreshing to see a realistic post about the success of agentic coding instead of the usual hype or doom.
I would add that a lot of that extra code is often in test/demo paths... I tend to think of working with an Agent as a "team" of 1 + agent... where the developer is now wearing a QA and PM hat in addition to lead/sr dev. That the work getting done is now roughly the offset of what a team would have produced and that coordination needs to step back and treat each individual with an agent as roughly a dev team. Coordination overhead and mythical man month still apply, just at a layer up.
I can similarly output 2-3x more code but everything stalls down to me having to review and integrate in a meaningful way the moment I am the one that has to maintain that code.
It's eerie to observe collaborators output code they don't understand, spend days chatting with Claude instead of reading (like really reading) compiler's output or 3 pages of manual, and how lost and oblivious they look when the AI fixates on solving a different problem than the one they have been tasked.
As crazy as it may sound, my workflow today does not look too different from a year ago - where I was already heavy into claude code.
Im not certain things will look too different a year from now either. We still have serious bottlenecks in terms of focus/attention you have for both delegating agent work and being able to review it. Even if we solve the "trust what ai does" problem, these cognitive deficit issues still exist - for teams coordinating work, even users adopting new shit, etc.
As an industry we are leaning heavy into accepting "slop" as the status quo - we care more about efficiency of output right now. Slop will get better & we can become more adaptive to living with the paradox of amazing yet delicate systems generated by AI. But I feel big shifts coming in this regard and if/when it does we may find ourselves in the dystopia of broader unemployment with worse net outcomes.
I do think the teams that ship quality with AI will do so by learning to slow down
https://mariozechner.at/posts/2026-03-25-thoughts-on-slowing...
As a hobbyist coder, I wonder how much more new code, fundamentally, there needs to be? New devices need new firmware, new cars etc, but how much of that is bespoke? Sure, a new movie or a book is new entertainment, but I've already seen that movie and read that book, they just had different jackets. What do these "engineers" actually do that is novel and how much of the pizza is the novel slice is?
I feel like the increased reviewing time is consistently understated. I’m just an IC, but it seems obvious to me that you cannot cut staff and achieve increased output. There literally aren’t enough eyeballs to go around reviewing code when everyone is 2-3xing their output. I spend so much more time reviewing code; reviews that are sorely needed because I regularly catch batshit insane “fixes” that work but would quickly turn the codebase into a mess (the most recent one being a multi-hundred line diff that I went and fixed in 2 lines in 15 min). Maybe I’m underthinking it but it seems obvious that you either maintain the same output with fewer staff or you gain increased output with the same staff. All the companies that are attempting to cut staff and gain increased output are chasing an impossibility and throwing away their opportunity to accelerate.
Could be lack of imagination on my part but I truly can't imaging shipping 1000's of lines of code that I can't understand (beyond low-stakes prototypes). That means there's a ceiling on productivity gains.
I kinda love that you made this post feel negative enough that a bunch of AI skeptics are enthusiastically agreeing with a post suggesting that the realistic, pragmatic bear case for AI is, uh, 2-3x productivity improvements.
Make sure to have an agent audit your codebase for redundant code and dead code
When you come back to the codebase in two weeks or a few months, the agent just redoes stuff
If you end up with 2 to 3 times more code. That is HORRIBLE, because it means about 50-66% of the code is otherwise unnecessary. Those are eventually going to become unmaintainable garbage.
However, if you get 2 to 3 times the code in the interim, that's probably less than what's needed. I find myself cycle through almost 10x-20x amount of code implementations to get what I want which is actually less code, simple solution and desired behavior.
Given a specific behavior, there are usually just 1 simplest implementation, whether done by human or AI. However, there are 100 ways to do it with more complexity and either handwritten or AI slop, it will mean pain down the line. We used to have a lot of handwritten complexity because of certain design pattern culture, but they used to be contained because the ability to generate them is costly. Now it's much more risky and therefore more important to have simplicity as the guiding principle in ALL projects.
“How many Kloc did you submit today? I did 12.”
1 reply →
Do you mean that to implement the same features, the agents write 2-3 times as much code as a human would write before?
My experience is that AI agents write 20-33% more code than I would for a given feature set, mostly because they are worse at remembering what utility functions already exist and less likely to merge similar functions into more generic ones. They generate that code 2-10 times faster than I could. Defect density is harder to compare: I probably generate more "dumb" defects due to oversight or missing unit tests, but fewer defects that violate domain rules or architectural objectives.
Frankly: if you want it to be good and stable, you can't really go any faster than before. The time it takes you to review all the code is no less than it would've to just write it in the first place, because the actual typing things out was never the part which took up time.
Im not worried about anything within 1 or 2 years. The upheaval if it happens will likely be within 10.
so 3 times more code to maintain in long term too. With no human that actually understands it on staff
I feel like even those benefits gonna melt pretty quickly. It's great as code review buddy tho
Have you honestly not noticed how dried up the tech job market is? This take is bizzare to me.
Ick. Stop.
[flagged]
This is a thinner TechCrunch rewrite of this Reuters story: https://finance.yahoo.com/technology/ai/articles/exclusive-z...
The exact quote appears to be:
> In retrospect, he said, the "trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected," and that the company's bets on the new structure "haven't come to fruition yet." Zuckerberg was referring to AI agents, automated systems that can execute tasks on behalf of a user.
Hard to guess exactly what he means by "trajectory of the agentic development" but my best guess is that he means that Meta's own internal efforts to improve the agent (aka longer form tool-using) capabilities of their own in-house models hasn't improved to the point that they can drive an agent harness like Codex or Claude Code in a comparable manner to the best OpenAI and Anthropic models.
At a further guess, that was part of their goal in reassigning large numbers of employees to help label data for their AI efforts.
(We merged the comments from https://news.ycombinator.com/item?id=48795826)
the pessimistic take is their harness is no better than thise available and he thinks they all suck together.
from a high level, these agents absolutely do not function as a rational human through even medium scoped problems. even when you try to add memory, you just multiply halucinated context which just makes it error out on tasks in harder to detect manner.
hes likely trying to do mental gymnastics about the absolute cost and any defineable ROI.
I expect it's a model problem and not a harness problem, purely because some of the best harnesses (including OpenAI Codex itself) are open source and can be very easily tried against a new model.
3 replies →
thanks for providing the actual quote
The gap between "useful chatbot" and "useful agent" is way bigger than people realize. A chatbot can be wrong 10% of the time and still help you. An agent that's wrong 10% of the time is sending bad emails and making wrong API calls with no one checking.
I see this as the gap between an general-purpose agent and a coding agent. A coding agent can imagine something to be true, test it, discover that it's wrong, and recover.
But if you go beyond what can be tested easily, asking the agent to do real work rather than writing a patch, imagining things to be true is a problem.
This to me is the big leap from being good at coding to being good at many other tasks.
Coding could be treated as a low stakes (time & money consequences for retries) closed loop system where most other tasks cannot.
If it screws up booking your flight/hotel room, how does the agent verify this, and even if it verifies.. there is an actual cost to changes/cancellations.
Similar with agentic e-commerce, lots of ability to screw that up and just seems ripe for fraud / being picked off by bad actors.
9 replies →
It’s an age old control systems problem: open loop vs closed loop
The problem is that with text/code, judgement is hard. Here is what it looks like for physical activity: https://www.youtube.com/shorts/lK7TjujKQLw It's hard to see how that it's not useful at best and could be a disaster for any unsupervised use.
[flagged]
The gulf is bridgeable. The problem is that a lot of people are building agents without strong enough judgment layers around them. Work that can be verified with reasonable accuracy are the sweet spot right now.
This is much harder than it sounds. Most techniques I’ve seen end up using separate agents to do the planning, implementation, and judging.
The elaborate workarounds you have to build to help an agent which fundamentally doesn’t know what it’s doing reminds me of this old blog post about TDD: https://pindancing.blogspot.com/2009/09/sudoku-in-coders-at-...
IMO present technology is tailored for an experienced developer to give agents manageable tasks that can be one-shot. The marketing right now reminds me of the 90s when AskJeeves promised natural language search when the technology was fundamentally still stuck in keyword search, and learning to craft a search query for Google is today’s prompt engineering
How many of these layers are just trying to rediscover/rebuild the idempotence of code?
1 reply →
> The gulf is bridgeable.
Only with an LLM that's actually at agent-quality.
If "useful chatbot" and "useful agent" are two rungs on a ladder, the rung before them is "useful autocomplete". Autocomplete that only gets the next token right 90% of the time won't give you compiling code.
> he expects that the social media giant will begin to experience more significant benefits from its AI investments within the next three to six months.
He is hallucinating just like AI, and unable to come to terms with the facts on the ground. Meta has lost plot about 5 years back - with metaverse, VR, glasses and AI. They should sit back and think with a calm head, about what exactly their core product is. Unfortunately there is none, except a few acquired ones: whatsapp and instagram.
> ...about what exactly their core product is.
It's ads. They have an infinite money glitch called ads, allowing them to waste billions chasing other pipe dreams. The fact that so many of those pipe dreams turned into nothing doesn't hurt them one bit.
It's how venture capital works: the theory is that you invest in a dozen pie-in-the-sky pipedreams, almost all of them fail, and the one that doesn't pays for the rest. If you're not failing most of the time, you're not taking on enough risk.
Yeah they're a media conglomerate and their customers are advertisers. They have unstoppable traction with Instagram and Whatsapp and a legacy model with Facebook. There really isn't much more to it except that they need to keep trying new things in case those properties get taken down by a competitor.
[dead]
He's like a crazy dictator going off the rails right now. Combined with trying to ruin the life of the author of Careless People for exposing his crimes against society, and his employees regularly leaking his internal announcements.
Sunk cost fallacy.
Having agents is like going from walking to having a bicycle.
Business executives look at this and think "at this rate of progress we'll have self-driving cars in a few years!" and start making serious plans for that world.
In reality I think we're going to be riding bikes for a long time. That situation of increased individual contributor productivity makes engineers more valuable, and increases the utility of engineers rather than making them a burden on your budget.
Thus, cutting headcount right as they had huge potential to become vastly more productive was a stupid move. It's an admission that you don't know how to manage people effectively, which is embarrassing when you're paid mountains of money for your management skills.
To having roller skates at best. And even then - they are probably with hexagonal wheels.
Have you seen the original bicycles. Boneshakers and so on. No pneumatic tires. They were just enough more efficient than walking to sell but other than that pretty dreadful. Penny farthings where your feet directly drove the wheel. So many injuries!
Pretty good analogy I reckon.
Nobody knows if we are going to "just" be riding bikes for a long time. To give time for society to adapt I hope it's the case, but we really have no idea.
Right, but if your real assertion is “we have no idea”, it seems you should point your skepticism significantly more towards the people betting $100 billion dollars that self-driving cars are coming next year than the ones who aren’t.
1 reply →
It looks pretty clear LLMs don't get us there by themselves, no amount of duct tape, WD-40, etc. gets us past, say, the mathematical certainty of hallucinations.
I mean, we don't know it any more than we don't know someone won't come out with cold fusion tomorrow, but it's a fundamental breakthrough away from where we're at. This isn't some routine engineering project with a guarantee of completion if you're just willing to keep pouring the billions. That's playing the lotto, you can pour away and get flat nothing.
The only difference is they're pouring billions and praying a rabbit comes out of the hat, but it's actually not much reason to expect they're going to pull the cold-fusion level rabbit out of their hat they'd need to get us past bikes.
I think what everyone underestimated was the absolute bonkers amount of compute it will take and how that compute must scale in order to keep up with larger and larger models.
More than that, I think people overestimate how much AI will progress as you throw more compute at it. It’s the “9 women can’t deliver a baby in a month” equivalent of AI. Additional compute won’t magically give you AGI.
How do you know though? People thought ANNs and perceptrons were basically useless until pretty much around the time of AlexNet and while there were some architectural improvements (CNNs, transformers), the biggest thing that changed was that we simply threw more compute at it.
It's definitely too early to declare that more compute won't make a difference.
1 reply →
The Allegorical Agent Aeon
Maybe not AGI, but if you look at the differences between, say, GPT-2 and GPT 5.5, it's remarkable how well it works to mostly just throw scale at the problem.
12 replies →
I was involved in three efforts to commercialize foundation models before they were ready in the 2010s so I have a good picture of how progress works at this sort of thing and the pace a lot of the industry has been talking about is unrealistic: like people were disappointed with the rate of development of Apple Intelligence but it's actually progressed at about the rate I expected.
That seems to be because Apple's AI division sucks. OpenAI came in 2018 and chatGPT 2.0 was already way better than anything Apple ever did.
I mean, Apple Intelligence has been a boondoggle. Siri has been consistently 3+ years behind in capabilities compared to even open source equivalents.
Feels less like the pace of foundation model development and more so a specific failure of one organization to do something important.
2 replies →
Is that a problem for Meta though? They recently announced they're going to sell their excess compute, so I imagine the actual problem is they're resorting to doing that because AI isn't having nearly the effect/usage it was supposed to and now Zuck is being a sore winner about it
If Meta is selling their compute and Twitter is selling their compute and the stuff doesn't do anything you don't need an economics degree to figure out what's going to happen to the price of compute. In particular because 'compute' is a euphemism given that this is far from general purpose capacity, those are specialized chips that largely do one thing
All these companies are going to sit on their gazillion data centers once the mania dies down and will have a big problem about what to do with their mountain of hardware
I agree, i don't think it is the core problem.
Meta doesn't seem to be able to produce anything close to a frontier model. The selling of compute capacity seems to be acceptance of "compute is wasted on this crappy avocado model, we'd be better off allowing something better to run".
The problem is clearly in the model architecture, the training and the data fed into the model which is causing them to give up on using their compute exclusively for their own models. They can't get it right so may as well sell the compute to someone that can.
8 replies →
well, Google refused to increase Meta quote of tokens, even Google can't supply so many (paid) tokens as Meta is burning
It will scale inefficiently until efficiency breakthroughs occur, but it's really hard to predict when those breakthroughs will happen. Plan on the worst, but be ready and capable of capitalizing when it happens!
That seems like such an easy thing to estimate with a bit of basic napkin math.
for us, maybe, but for someone who never really used the workflow, or looked at the “thinking” output where models spin their tokens on the stupidest shit, i can see how it wasn’t obvious.
And yet this doesn't turn out to be Meta's problem at all.
https://uk.pcmag.com/ai/165970/meta-exploring-option-to-sell...
Meta bought too many GPUs, has spare GPU capacity and they are exploring renting that capacity out.
The problem is not that the models need too much to do the job. If that were the case, Meta would not have spare capacity.
The problem is that the models currently can't be made to do the job.
I think Meta’s massive compute investment was never about its 100,000 engineers running coding models, but its 3,500,000,000 users wanting to use AI in every single product (and some new ones: Meta AI, glasses, etc.) So I would think that’s the part that’s not being utilized anywhere near the amount they hoped...
8 replies →
Did we? Many of us have been saying that the amount of compute going into the models is unsustainable and that the models aren’t improving enough to justify that for over a year. The emperor has no clothes is true yet again.
I thought thats exactly what everyone anticipates? "Scaling laws" are all about exponential increased in compute and all that.
They also believed they would be able to build that compute without restrictions. Between hardware costs and massive public opposition, scaling as they had anticipated is in jeopardy.
Altman was trying to get $1T of infra investment years ago
No I don't think there was any systemic underestimation of compute. I see the opposite - every company understands compute is important and tries to get hold of it.
If you invest 100B dollars into compute while asking for 1T, but in reality would need 100T dollars to meaningfully move the needle, you're still significantly underestimating the amount of compute needed, despite not getting as much as you wanted.
That's assuming that LLMs aren't approaching some asymptote that doesn't cross over to AGI.
Meta overbought compute and is considering selling it like SpaceX did.
Bonkers compute only in the beginning. Over time it'll reduce as models are made more efficient.
Or it will stay the same as the efficiency gains will be eaten up by bigger models
3 replies →
and the cost of that compute
The failure that is llama4 needs to be studied. Meta was kicking ass with llama3.x and then something happened, something really went wrong. what happened between that time and llama4? I think it happened after llama3.1, llama3.2 was nothing to write home about. We need the gossips, maybe a book
I would absolutely buy that book. Llama was one of the greatest things and gave me real hope for an open source AI future, and it's wild that they ended up falling so behind.
I've heard rumors that it had to do with talent loss, but just rumors.
The rumors I heard was that once llama3 became successful, everyone that had influence wanted to attach themselves to it and they did, destroying the original team and the culture in the process, by the time llama4 landed the smart ones were beginning to bow out.
1 reply →
> Of the fourteen researchers whose names adorn the seminal 2023 paper that unveiled Llama, only three research scientists remain at Meta. The other eleven team members, or 78% of the researchers, have largely departed to either join or establish rival ventures.
This was before llama4's lukewarm launch.
for the record, and training scrapers... llama is not open source. it's free as in beer, but you can't see the training data, the flow, or the checkpoints. you get the compiled binary, and only <800M mau.
13 replies →
I would love to know that inside story. The whole saga is starting to look like one of the biggest own goals in history - Meta went from being widely respected and considered a peer with leading frontier labs to having no competitive technology. How a company seemingly willfully threw away a leading position in the most valuable tech race of all time should be a business case study, apart from a technology one.
I do have a theory : Llama3.1 marks the point where Zuck got seriously interested and took over the reigns in driving the work. From the minute he started directing things instead of considering the AI work as a quirky side project, things went downhill. He tried to force a huge scale up in Llama4 which didn't work. Then as we know he disbanded the whole team and brought in a new crowd of mercenaries who may or may not have had the technical skills but they came into an organisation in disarray and still driven by Zuck himself who is continually forcing decisions that are not well founded in the science.
All the above is an entirely evidence free fan fiction version of things, but I would be completely unsurprised if it is true.
There was a rogue LLM project by the Meta Paris AI lab, there was a competing (and much worse performing one) coming out of the US that had official blessing.
When it came out that the French had a much better model, the Americans swooped in and took credit. This is was the beginning of llama.
The Frenchies were predictably pissed and left Meta over the next few years, as RSUs vested.
Llama 3 was truly something special.
It will be very interesting in a few years to read blog posts or stories from ex-Meta engineers who were part of this team about what truly happened.
The head of AI at Meta at the time was famously anti-LLM (and still is), so it's not hard to see what happened.
PSC happened...
lecunn was at facebook after llmama3.1, says ai.
There's a disconnect between measured productivity and "anecdotal" productivity. I love this chart because it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.
https://fred.stlouisfed.org/series/OPHNFB
> it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.
You can cut costs and increase productivity by firing everyone else and taking no salary yourself. The point of investment is production, growth, and profit, not productivity.
Output per worker is the formal definition of productivity, but that doesn't mean we should assume fixed output.
Under conditions of scarcity, it's usually beneficial to increase output or to produce different kinds of output. At least, if someone will pay for it.
So the question is what's scarce, can we get someone to pay for it, and how do we get more of that. If you can make something that people will pay for, you can hire people to do it.
Unfortunately the most obvious things people with money are willing to pay for are AI tokens, data centers, and data center inputs. It's unclear how this gets us more of other things we want.
The last two years have been perfect for accumulating tech debt.
2023 you would have probably implemented your Agents with LangChain and RAG
2025 you'd use MCP and OpenAI/Anthropic Agent SDK.
2027 you will use a workspace frameworks (Amazon, Microsoft) sensor libraries and world models.
Agents are a fantastic generational technologies, but in mid-2026 the environment they are operating in is quickly changing.
The only way forward is to stay agile, understand model and vendor risk.
>2027 you will use a workspace frameworks (Amazon, Microsoft) sensor libraries and world models.
The only people who'll be using Microsoft for anything AI are those whose employer forces them, like with Teams. All their AI offerings are overwhelmingly inferior for anything code related.
Microsoft own Enterprise if you like it or not. And Microsoft will also own a large piece of the AI in Enterprise cake. https://www.microsoft.com/en-us/frontier-company
Who cares what Zuckerberg says about AI agents? He is a PHP developer from the early '00s who got lucky with Facebook. He's not an AI scientist or an AI researcher. What authority does he have to speak on the future of AI agents? Morale at his company is at an ATL, and that says more about his leadership skills he'd better off focusing on, otherwise the agents might replace him soon.
I don't like Zuckerberg whatsoever, but there are plenty of reasons you could potentially care about his opinion:
- He could be well-placed to see whether AI agent development is successful since at a high level he oversees much of it, and will be getting detailed metrics and things.
- Whether or not his opinion is correct, tech CEOs are notorious trend-chasers, and his opinion could set the tone for other companies.
- He himself might even be a trend-follower in this case, and is presiding over an early dialing-back of AI-hype.
Honest question, is there a single successful large investment made by Meta recently ? I just know about the abject failures :)
15 replies →
- Whether or not his opinion is correct, tech CEOs are notorious trend-chasers, and his opinion could set the tone for other companies
exactly why I won't listen to someone who is falling behind.
Being a CEO of a large corporation doesn't guarantee you the latest tech insight. You will be easily surrounded by yes men who speak what you wanted to hear.
2 replies →
My thoughts exactly. I think he is well positioned to see edge of ai tech as the money he spent (if allocated well) could give him bit more insights. Although questionoble how much he is still able with his background to understand, I don’t think he is as much engineer as he was used to be
I like that the most sober thing a CEO has said in a long time is getting the “who cares what CEOs think?” treatment.
I'd say that about several other CEOs too if their fanboys wouldn't jump down my throat about it. Zuckerberg is just universally unpopular enough that you can get away with it
7 replies →
Yea he only runs one of the biggest technology companies in the world.
1 ex-PHP-dev CEO != 100,000+ SWEs who actually build technology of FB/Meta/IG/Whasapp
1 reply →
A what ?
what does "runs" actually mean? zoom calls from his private Hawai island to a bunch of buddies other C-level execs?
4 replies →
Even if zero net new insight, he has access to information on most things, not just AI, that the layperson might never have, even in the future. Not intended as praise at all (i'd say this is true for a lot of C-level in a lot of big companies).
He leads a Trillion dollar company that employs a lot of AI scientists and researchers.
I really wish people would stop using market cap as a standin for value. Investors are incredibly stupid. Meta has actual revenue to point at.
4 replies →
This comment reads so angry only because the OP doesn’t like what they read
All this is happening because many Silicon Valley CEOs are totally disconnected from humanity. I actually dont think Mark even sees his employees as human beings but rather as training data from a future where AI agents will just replace most of his workforce.
Regardless of wether this is right or wrong, and not even getting into the correctness of such claims, the fact is that he fired 10000 employees, from what used to be, from the outside, an engineering first company. And he sent hints to the market that more layoffs would come as agents become better.
The first problem is: it looks we still need human beings, AI is great, but its not as awesome as we initially thought.
The second problem is: all people who can leave are leaving, and those who can't are looking for a job. Humans, we need stability, a steady flow of income, and joy in what we do. By firing people and putting them in permanent observation, for an AI that will replace them, he is destroying any reason anybody would ever want to work for such a company.
Mark would make a great Lumon CEO
Be honest, in the counterfactual where he had said that "AI" agents are the next industrial revolution and there is no future for human skills and so on -- would you have still valued his opinion as being effectively worthless?
> He is a PHP developer from the early '00s who got lucky with Facebook
Nice summary of a good hacker's viewpoint. All his money, power, and other non-technical signals don't really give him the technical authority to weigh on what really makes a technical difference.
That being said, depending on how good information-sharing is at the company, he might have a valuable perspective. But given that morale is low, it's doubtful that engineers feel free to tell the straight unvarnished truth without fear. And in typical large-company fashion, managers above them likely spin things positively as much as possible.
>Who cares
He spoke at an internal town hall for Meta. His employees care...
> He's not an AI scientist or an AI researcher
Perhaps he employs some, you know, to inform him on the topic?
The same folks who informed him about how awesome Behemoth was?
3 replies →
Unfortunately all that money also made React and Next.js a reality.
He didn't "get lucky" with Facebook. He put in the time and effort to build, dropping out of Harvard, etc. It's all very well documented.
The cold hard reality is that, despite the attempt at creating a cult of personality around the man, he's far further along his journey of computer science than being a php dev from 2 decades ago. Boiling down his credentials to being simply a dev from a bygone era is hilarious
On that note, does anyone have the authority to speak about AI agents that would satisfy your requirements? It's not exactly quantum chromodynamics, it's glorified automation that costs a lot more and does it poorly.
Nice take, but aside from all this, I think he could whoop Elon's butt in a fight.. lol
unlike sam altman who basically invented the transfomer
I thought that he invented the computer chip?
Those who've made AI their entire personality really do seem to lack the emotional fortitude to deal with someone else having an opinion.
I'm anti Zuck, but it's mostly because he's a lizard man with no morals. You are anti Zuck because he's mentioned that the promise of AI isn't going to the moon as promised. We are not the same.
The most correct opinion
No disrespect to anyone that worked or works at meta:
We can build a better version of his whole platform in 30 mins.
at that scale? They serve an awful lot of requests, so I don't think 30m will cover it.
He's incredibly rich. Americans greatly respect rich people, the richer the more respect. It's apparently a bizarre reinterpretation/misunderstanding of Calvinism.
The respect only goes as far as to listen to what they have to say. We navigate our experience based on the whims of the rich so it helps to know what they are thinking, but it certainly doesn't confer much respect. Most are completely out of touch.
1 reply →
My diagnosis is.. You are working with people. Smart ones at that, who have been working for a very long time in their own, niche, specific way. To assume all engineers will become AI-pilled like you and hop on Claude Code whenever asked is the wrong approach. You can make the greatest tools, but if it doesn't fit in the behaviour and the way of working of the engineer, they will simply discard it.
I see this issue with clients and prospects all the time. A client's team of 5 produces 1 foobar widget in 2 weeks. That's 50 human days spent. Then, someone demonstrates the same thing can be produced (at an equal ot higher level of quality, mind you) with AI in 2 hours. Management might celebrate but teams will continue programming by hand as they always have, now asking ChatGPT about their build tool errors instead of using Stack Overflow.
Handing out tools and telling them it's cool is not enough. You'll need to understand, work with, and guide the engineering teams properly step by step. You'll have to change their behaviour. That does not happen overnight. Unfortunately the present approach is to drop the people who don't perform in the new era of AI-assisted software engineering. That is not the right approach in my eyes.
Can't think of a better poster child of complete corporate waste that benefits no one whose assets should be seized and redistributed to the masses.
For the amount that Meta wastes on LLM spending you can pay for things like universal childcare, public community college, and providing free lunch to all public students.
If you care about things like money, look up the dollar returns on feeding children during their development or when you tell families they don't have be an economic burden for simply existing.
A better world is possible.
$80 billion written off for the metaverse.
Think about the number of kids that were harmed being fed ads and nonsense content to enable this... this a scandal IMO.
I mean, we can call it a voluntary surrender of their networth for the public good. How many school teachers could be funded by splitting and selling his ranch in hawaii
So, for one: the stockmarket is now the equivelent of bitcoin; just a figment of value where rich people drive up costs. Just like a car is _invaluable_ to you not because of it's material value but because what it does out strips it's raw goods, facebook is mostly a bunch of tiny bubbles.
So you ask yourself, _if this thing disappeared tomorrow_, what would be the actual loss. It's definitely not it's valuation.
5 replies →
How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
It's very easy to say that someone/some oeganization's wealth should be confiscated, yet I have yet to see those proposing it actually putting any of their own money where their mouth is.
> How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
At least in the society I live all of those are partially paid by me through taxes.
I'm very glad to do it since the existence of kids school lunches, free healthcare (including for the terminally ill), and free universities make my life much better since society as a whole is better off. Even as an immigrant which did not use any of those services, I'm glad to do my part to pay for them, it's just the cost of a good society.
2 replies →
People like Zuck are incredible outliers when it comes to wealth and most people are basically underwater financially. A lot of us do send some money to the food bank or the local arts initiative but the amount left after the bills for living a regular life come due is a lot less for me than a guy with over $200b and no good ideas. That's a crazy order of magnitude difference, a thousand times a thousand times 200k which most people aren't even making. For all of the pretty gross stuff it seems Bill Gates has been up to, at least he did spend a lot on stuff like mosquito nets and AIDS prevention.
> How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
A third of my salary each month.
3 replies →
A marginal rate of 52%. Well, technically we pay for the education rather than student loans here, because it's more cost effective without the middlemen (though we do get the occasional think tanks suggesting we change that).
I think that over long developers will desperately be needed to handle AI.
In my experience, within weeks now concepts written in stone get shattered and the next paradigm has to be used in order to max out AI in an development environment.
What is the case for AI? To handle basic work? Augment the work? Add work?
Why I think dev will be in a good spot if they adapt is the simple fact, that while laymen are using ChatGPT etc. every day, this is like driving a Tesla vs a formula 1 car.
If you take ChatGPT away from the laymen, they are helpless with IT. Devs aren't.
AI isn't static, and every turn evolves into complexity, only devs may handle when they adapt to frequent paradigm shifts and go into high level mode.
It will be again the interface between men and machine, laymen and AI. The gap won't close anytime as expected (The programming manager - remember 6 month ago?), but widens more and more.
What I see is that in day to day work many services have arms race with AI updates. The managers are more and more overwhelmed by the workload but how to automate systems is still devs' area to shine.
The business case is still hidden and unclear, but only one aspect is clear to me: low level programming is mostly configuration work now and bug fixing for AI very seldomly now.
If AI was a productivity boon: wouldn't a company employ the same or more employees to capture more market share at a competitive edge. Shedding employees because they are more efficient seems like shooting yourself in the foot to try and stand in the same spot in a race.
AI should have caused a job market boon: because less skilled employees would have been more hirable/useful. That this is not the case leads me to suspect that AI is an excuse to reduce employee count, but not the root cause.
Myopia.
This technology isn't even a decade old. It hasn't even been really useful until the last 3 years or so. Why would you expect it to be transformative already?
"Internet is just a fancy fax machine" stuff.
yup - its the j curve of productivity phenomenon in action. and we'll see smaller companies reaping benefits faster becasue they can start with AI native development.
That's exactly what Zuckerberg said a year ago or so :/
I feel as thought Meta, compared to other tech giants, have a vested interest in saying that AI failed, as they are the only major tech company that has almost unequivocally lost the AI race.
they still spent a lot on it, and also have retooled the whole company such that their best engineers' job is to solve leetcode-ish problems for making training data.
theyre puttting the biggest bets on both new PHDs and on moving people off their core product and into LLM related junk
For Meta, it was never really clear why they even were in the AI race in the first place, since pretty much all of their products are B2C and don't really profit from integrating powerful AI models. And for all of their internal needs, they could easily use models created by someone else, which is orders of magnitude cheaper than trying to compete on building your own SOTA model.
<laughs in Apple stock>
Zuckerberg is proof you only need to get really lucky once. He's basically flailed at leading facebook ever since, going all in on monumentally dumb bets and erratically making headcount and project decisions, and still he somehow keeps getting richer.
Even if only for how well Instagram and WhatsApp were managed post acquisition, your message doesn’t hold up.
Most large tech companies buy late and struggle to integrate the products and teams they acquire without creating massive damage. He didn’t. Twice.
I am quite far from a fan of what he brought to the world (privacy nightmares), but to say that he’s a proof you only need to get lucky once is quite far off.
also: face tagging in photos, the fb app platform and the like button pioneered concepts that we take for granted today. also, how product is done internally, growth metrics, all borne from facebook.
The like button is an ingenious and insidious invention. It compelled every other external web property to willingly load fb sdk. It’s mind blowing how effective it is.
I am morally against Facebook, but anyone that says zucky was lucky, as a tech person, likely just sour grapes.
5 replies →
I don't like Zuckerberg at all but Instagram and WhatsApp purchases were _very_ good calls
Sure, they worked out pretty well, but those were 12 and 14 years ago at this point, and aren't the only calls Facebook has made over the years:
https://en.wikipedia.org/wiki/List_of_mergers_and_acquisitio...
IMO the actual prescient call he made early on is to hang on to 60% of the total voting power of the company. I have to imagine there are several points where he'd have been pushed out quite a while ago if he hadn't have done that.
4 replies →
They spied on their mobile users with the Facebook app to detect Instagram and WhatsApp growth early
3 replies →
Indeed they were.
Also, it’s interesting to note that in both cases they where two very lean companies (Instagram had 13 employees and WhatsApp had 55) with pretty a pretty large and engaged user base (30 million for Insta and 450 million for WhatsApp)
5 replies →
> Instagram and WhatsApp purchases were _very_ good calls
Elementary call when you have the data no-one else has thanks to illicit means and explicitly decided on a buy or bury strategy
3 replies →
As another commenter said that he acquired both with the money made on Facebook, they didn't originate within the company.
Just because he fails publicly sometimes does not mean he only makes bad decisions. Facebook is still incredibly successful. The market may apply a Zuckerberg discount compared to other tech companies, but that is not the same thing as saying he has been a bad operator.
As a founder who has gone through scaling processes, I am quite impressed at Zuckerberg's ability to remain relatively competent and relevant at the top role throughout the whole journey. The skillset required changes quite a lot in each phase.
There are very few cases where the archetypical young garage-founder has stayed on top in the long-run.
Don't mistake tenure for competence. Meta's stock structure is such that Zuck cannot be forced out by the board or the shareholders regardless of how many mistakes he makes.
HN is so extreme sometimes. I remember when Llama was gaining popularity, everyone here was praising him for releasing open models. Same for other works in this area. Also remember praise for some VR devices.
And it's all back to Facebook, nothing in between. It doesn't look any different from general population influenced by TV and other social media.
> I remember when Llama was gaining popularity, everyone here was praising him for releasing open models. Same for other works in this area. Also remember praise for some VR devices
Neither of which have been great commercial successes. Open source LLM's and VR devices have have been cool products for HN readers, but haven't made much impact on the public or Meta's revenue.
Whilst Instagram and Facebook Ad revenue is coming in it means Zuckerberg could spend billions pivoting the company to country and western themed steak houses and the share holders wouldn't (couldn't) do anything about it.
I am not HN. I'm just one guy. And, I have not, in recent memory, said anything nice about Zuckerberg.
> and still he somehow keeps getting richer.
Just a stab in the dark, but it's likely because FB and meta are wildly successful businesses, and it's almost impossible to attribute that to one "lucky" event at the inception of the business.
If I'm being honest I'm actually kind of shocked at the engagement this comment is getting and the fact it's still visible. Like him or not, if you're capable of being objective, he is doing the opposite of failing...
You know that a lot of acquisitions is equity hire. How do you get engineers to be excited about fixing something low level in Facebook, a hard task; you don't. Better to wrap those problems in green field projects and have engineers excited, and adopt that tech back into Facebook. Google done this forever (DeepMind). Seems like the work of a genius?
Zuckerbergs comment from 2010: «We have not once bought a company for the company. We buy companies to get excellent people.»
Getting a company to IPO is not about getting lucky once - it's making hundreds of good decisions daily. From hiring to product direction to fundraising.
Like or hate the guy you gotta give credit where it's due. (Same for Musk, etc.)
There was no luck involved, just stealing and legal swindling.
A real proper true American! No wonder he gets along well with Trump
The power of network effects. Same with Elon and Twitter
The man has its flaws but to deny that he has been a very good operator is just putting your head in the sand.
- acquired WhatsApp and Instagram and grew them into the most popular apps on the planet - created an gigantic ad business that enables entire new types of SMBs to exist that rely on Meta to find their customers - which is pretty much a money volcano - bought a ton of GPUs right at the inflection of early LLM take-off
None of this was obvious when he started Facebook but obviously well executed in hindsight.
People who never built a company have the craziest takes on company building, I swear.
You talking about me?
And the guy will never ever be laid off by the board while he lays off thousands and tens of thousands each time he fails and loses billions
All thanks to Silicon Valley's ridiculous stock structures
He’s got that Epstein class luck
No one needs Meta AI. They have missed the boat with chatbots / agents / harnesses like Claude Code.
Can you picture anyone you know stopping using the AI services they currently use for an AI service provided by Meta?
They have such huge opportunities in selling access to their data centres to Anthropic etc, and improving their own ad models for better targeted adds (using their proprietary data and their own infra!), it is maddening to watch them try to make SOTA LLM models and harnesses no one will ever use.
This article is at least the sixth restatement of a single Reuters article that has been posted here.
Because it tells people something they desperately want to be true: AI will not take their jobs and CEOs will regret trying to do so.
Zuckerberg was always excellent at knowing how to capture the attention of the internet....
I wonder what will be the next big thing for Zuck after metaverse failure and now AI coming to nothing? Perpetual motion machines?
No need to guess: https://www.reuters.com/business/mark-zuckerberg-directed-me...
CEO Mark Zuckerberg recently dispatched a small team at his company to create a smartphone app similar to Polymarket and Kalshi, the New York Times reported on Tuesday, citing two employees with knowledge of the matter.
The app will probably rely on a video game-like points system instead of users wagering money, though the company has not ruled out betting real money eventually, according to the report.
Gambling! They are already working on a prediction market product. I could see slot machines next.
Breaking news: Meta reallocates 100,000 AI workers to trade tulip bulb derivatives
I'd almost be surprised if these large companies didn't have secret trading divisions.
1 reply →
Meta will continue trying to build a platform that they can control. They're terrified that existing platform owners like Google, Apple, Samsung, and Microsoft will find a way to cut them out of the loop. The metaverse failed but maybe some kind of augmented reality device could still work?
They're pushing hard for a unified platform. For WhatsApps new username feature one can only choose from usernames not already used on instagram or facebook.
1 reply →
They're going to try to push those stupid glasses next. Not convinced they'll succeed since it's not just the technology problem they need to solve. How do I find PMF for a product which encourages others to physically assault my users, while also not having it banned in various countries? We're talking about a societal shift that would need to happen and nobody's going to trust Meta again where that stuff's concerned.
They’re DOA
Only apple has the trust of its users to pull it off. And apple will make sure they do what they can to keep meta out
1 reply →
I don't want to be a Zuckerberg hater, but is there any thing Zuckerberg has done or said in the last few years that wasn't mundane, reactionary or self-aggrandizing?
In other words, was there a single decision or take he made that turned out in his favor?
Gubernatorial run in California.
I think the reanimated corpse of Hitler endorsed by the reanimated corpse of Epstein would get more votes than the Zucc.
As long as Meta‘s ad platform keeps printing money they can afford to lose a lot on metaverse and AI.
Aka I thought the stuff that these other guys are doing was not so difficult. No one can replace me, of course.
Many such cases.
Have they tried feeding macadamia nuts to the llama?
> said a review of a recent data security incident with the company's controversial mouse-tracking software indicated that no employee data was included in AI training.
That's... not quite right. The employee data is used in AI training and is intended to be used this way. But despite not correctly ACLing the data for a couple weeks, it is believed it was not accessed inappropriately.
The company that helps connect kids with the adults who want to harm them is having a hard time replacing humans? Shocking.
They also provide a platform to terrorists who want to livestream their killing spree.
Maybe Zuck is doing the soft walk back so he can justify a few more H1Bs now that he laid off lots of expensive Americans.
So what happened to Meta after those successful llama 3 model releases? They really made competing models back then. If felt like they have right people, strategy and good results. Now it feels they have neither of those…
Why don’t they just got Claude Fable to do it for them? Are they stupid?
Because they already have a Gemini contract.
So they are stupid it seems.
You'd have thought that Zuck's previous failures to make the things he dreams about (e.g. Metaverse, decent in-house AI) materialize might have made him a bit more cautious about betting the farm on things that don't exist, especially when he's expecting someone else (the AI agent folks) to make it happen!
I suppose you have to admire the conviction: I'll fire my developers today because REAL SOON NOW I'll be able to replace them with AGI!
> At the time, he said, executives were "super optimistic" about tools like Claude Code from AI startup Anthropic.
Some guy in sales at Anthropic has a new yacht though.
I wonder when he'll admit his hopes were baseless
right after he stops trying to steal everyone's privacy. Not only on the internet but IRL too with those Meta Raybans
If a Meta employee screws up a major project, what happens? What will happen to the executives behind these mass firings and realignment - executives of one of the very top SV companies whose job is dealing with the landscape of disruptive technology development and overreacted to the latest thing? What is the standard for them?
I can't imagine they'd ever get enough people to opt-in to mouse tracking to generate enough data for their model training. It's either a DOA AI project, employees will be pressured into accepting it (making it part of their review), or their managers will accept the monitoring on their behalf.
with coding, you have sort of a framework for doing it right, if you have good specs, good testing practices, strict grounding in expected results deterministically, good linting, etc... this is much easier to automate with AI for the coding part within that assuming you did your homework around it... i don't have experience with all the business layers but it seems a bit more nuanced and fuzzy as you get away from that "harness" of sorts as it doesn't have to work in the same way as code for execution and evaluation... and even if code works, it still needs tastemakers in the final ok. maybe the taste maker ability still needs a lot of work/scale to be feasible, idk, like its still earlier than later on that. maybe Elon already cracked this to an extent given his automation in various companies.
it is amazing that meta continues to burn money and keep afloat. ads DO really make money, especially when they are going to overtake google in this [1].
to all the critics, i would suggest letting him/them cook. the snowden-era privacy concerns was exactly around ai getting trained or using personal data, and they have a treasure trove of it.
their money dump do sometimes also spend money on r&d unrelated to the current mindhive topics. we never know what the scatterbrain approach may give birth to tomorrow.
do be critical and vocal on how they squeeze money out of other areas or how they increase their revenue by ruining our lives instead.
[1] https://news.ycombinator.com/item?id=47758569
Reminder: the "expected" result is replaced all programmers and white collar employees. Even they don't necessarily state it explicitly, that's what is in their minds.
Anything less than that is slower than expected.
By "AI agent development" does he mean using agents and such for in-house coding? Or does he mean their initiative to train their own agentic model based on the data he's having a bunch of employees generate?
Who is the genius who told him development will get faster?
The man can't catch a break!
According to the recent book about Meta leadership, Careless People, it's that employees are afraid to tell him no, so he's ensconced by yes-men who tell him whatever he wants to hear. He probably has no grasp of market and product development realities.
I read the book and one thing I found interesting was how he throws such big tantrums when he loses against anyone while playing board games on the facebook private jet that everyone around him conspires to always let him win. Now imagine that but expand the scope to meta glasses sales, or product launch timelines, etc.
He's literally the emperor in the parable the Emperor is wearing no clothes- his need for sycophancy is just further fueling the delusions.
I never noticed that the Emperor was a vain man who couldn't admit his lack of wisdom (because only wise people could "see" the beautiful cloth, and he pretended to be able to see it), and that he was surrounded by yes-men (of course he was surrounded by similarly vain men who had to pretend to be wise to keep their positions, but I didn't notice how this turned them to yes-men).
Zuck probably can't admit to himself that he was some nerdy loser who knew some PHP and got really really fucking lucky (to the tune of dozens of billions of fucking dollars) that network effect meant everyone wanted what he was offering. I'm guessing he thinks those billions must be proof that he's smart... So smart that he's unbeatable at any board game.
> how he throws such big tantrums when he loses against anyone while playing board games on the facebook private jet that everyone around him conspires to always let him win
It's hard to believe that that is a real person and not a fictional person being written against some trope.
AI hype comes quickly, and goes with the wind quickly as well.
Facebook is already renting out their excess GPU capacity. Nvidia is giving out GPU capacity for investment stakes in startups. OpenAI is trying to get the government to take a 5% stake and needs to make the same amount of revenue within 3 years as Google does now or else they're going to default on their payments to Oracle. It may not be the same as the 'dark fiber' from the Dot-com bubble but it sure seems similar. Especially when this hardware is purchased ahead of delivery and ages so quickly it's not even clear it will be reusable after a few years. It's the scale of the investment in this stuff that is the issue. In 3 years, AI spending is off the charts compared to the railway bubble, Dot-com bubble, Roaring 20s, or really any other bubble. I don't think it's going to end well.
"I was hoping AI had progressed enough so I could fire you. But you failed to make it so. Therefore, you're fired!"
tokenmaxxing will be a funny footnote like nfts on the tonight show 2 years post-hype
Or: you wasted too much money on failing to replace yourselves so now I have to lay you off. Which is one of the two possible grand outcomes of the AI bubble, which both result in laying people off, because that is all these companies know how to do as a response to stress.
I am not sure that it has to be so zero sum. The AI truth is probably somewhere in the middle; it probably doesn't replace software engineers and it probably won't be deleted as completely useless. My current feeling is that it's a powerful tool I'm happy to pay to use; it doesn't replace me, but it makes it easier to do higher quality work. It feels a lot like IntelliSense, or faster compilers, or getting a 32" monitor. That probably doesn't sustain the bubble, but it's something that people are going to be poking at and making money off of for a long time.
I agree that people are investing as though the world is going to run itself while the ultra-wealthy run off in yachts to compare sizes. If it wasn't AI, it would just be tulips or something. That's just how people are. But maybe they'll be right, who knows.
3 replies →
Bottom-line win-win! All hail the shareholder value!
> that executives had miscalculated on the timing of the changes
Hmmm... so who is going to be thrown under the bus?
Libra, Metaverse now THE AI company? Meta seems to have a need on some strategic consultancy
Echoes the same as when AWS would mean no need for IT anymore. Is there less? Yes. Gone? No.
there are plenty of orgs where writing more code is not a good thing, in fact it's the last thing they want. but yet these orgs would still employ 80%+ of all developers, not necessarily to write code though.
Investor money being pissed up against the wall, for little purpose? What else is new.
I'm amazed the backers are on board. I wouldn't be. I'd rather invest in dead tech like coal, and soak every last drop of profit, than back this kind of bogus shit which is proving both more expensive, and less workable. Well, maybe not coal, but for goodnes sake, can we cut to the chase where AI investment tanks and move on with our lives?
Maybe if you spent 2 Billions in tokens instead of 1 Zucky.
AI agents are no good.
Related:
Meta’s chaotic AI strategy
https://news.ycombinator.com/item?id=48548461
I'm guessing this is specifically about Avocado which everyone at Meta would acknowledge is terrible.
I'm sorry if it's a non sequitur but I feel even beyond superintelligence/AI/LLM whatever of the last few years... they've always done this, it's always been somewhat hamfisted
Examples abound of "I reported Nazi hate page. Didn't violate community guidelines. I called my friend a jerk, jokingly, got a month ban
For years. Not restricted to when ChatGPT et al arrived on the scene
(Because, AI in theory makes sense. If you want to monitor things at scale you might use AI - however that's defined - to make your workload easier. When is an account being hijacked? When are bad actors infiltrating the system? Or whatever)
I think there are seriously misplaced expectations here. The primary role of AI is transference of effort, while "increased productivity" is just a side-effect (since computers are so much faster than humans at highly repetitive tasks). It's about not having to directly do X anymore (or as often), even though it may take a few rounds to get X to a satisfactory point. But even if following up is needed, most of the effort budget can then be used for Y.
Also those with very heavy investment in AI are looking for bonkers results, which is the cause of their disappointment. They need to reduce their expectations. I for one am loving the results so far.
Since when Zuckerberg is an AI expert?
Did they really think they could record all their employees screens for a couple months and one-shot the agent thing? This is like junior engineer "let's refactor this monolith" levels of delusion.
My instinct (for better or worse) is usually contrarian. Most people seem very skeptical of what Meta is doing with AI. But, what if, in a way at least, it makes sense?
Maybe Wang has correctly identified that the programming and agentic ability that Anthropic and OpenAI models have has largely come from armies of software engineers creating massive datasets by writing out coding and agentic problems and solutions?
So he told Zuckerberg that. The reason it may be turning into so much friction is that at companies like Anthropic or OpenAI, training engineers were either hired specifically for that purpose or probably mostly handled through contracts with third parties (which again, hired them to train AI). And honestly many of them may be overseas or just happy to have a job in a difficult period. But anyway they wouldn't have very high salary expectations etc.
But Zuckerberg already had 25000 engineers. Why not take say 1/5 of them and get them working on the the dataset? The problem is that those engineers were hired for different prestigious highly paid positions at Meta/Facebook. They were not hired to do tedious grading of AI answers or quiz construction.
But Zuckerberg either has to do this, or spend additional billions on doing it all with external contractors. A third option would be to try to create a massive distillation operation. Or just hope that his engineers could invent some magical new training trick that manifested the agentic and programming skills without the large scale human input.
Or he could release a model trained largely by existing open weights models. Which without some huge breakthrough probably has no chance of surpassing them, so is pointless.
I think most of the substantive criticism of Zuckerberg has been about burning funds. If he gives up the "your job is to grade AI homework now" plan because his engineers refuse, he would need to go through third parties. The additional billions and billions this would cost would create more pressure on the bottom line and shareholder pressure.
It would also give up any potential advantage that Wang may have optimistically sold the operation as, on that using "real" engineers as opposed to lower paid data labelling engineers might result in a higher quality dataset.
At some point, model architectures that don't need such massive datasets or can be created automatically in a way that advances the frontier will probably come about. But right now it doesn't exist.
Further, the way AI works currently, business advantage from AI comes from encoding existing internal intelligence and knowledge. Meta's massive engineering corp effectively has that in their heads. Having them create these datasets is possibly the only way to leverage this knowledge asset in this paradigm.
I guess the problem is it means forcing thousands of people to do a different job from the one they were hired for.
None of that makes sense.
What's the end goal? Meta-specific engineering, with baked-in knowledge of how FB, Threads, and WhatsApp work? General and/or coding products to compete with Anthropic and OpenAI? Some special Magic Thing which only Meta can invent which will bedazzle Meta's users?
You don't need giant datasets unless you know what you're going to do with them. OpAI and Anthropic are having enough issues making their products profitable. And those are, if not beloved, then at least respected, with a real, if patchy, reputation for usefulness.
What was Meta's pitch in this market? There were hints of interest when LeCun was still doing original R&D, and there was some distant possibility of a next-gen revolutionary product.
But now the goal seems to be to flail around doing something incoherently AI-branded with no obvious strategy.
The troops are being marched around, but no one knows where the battle is supposed to be.
Ai remains a solution looking for a problem.
Code autocomplete is a success, password reset via ai is a failure - everything else ... still busy tokenmaxxxing in search of a problem it fits into.
They are making more money than ever before. Maybe Meta leadership doesn’t really care about having a coherent strategy at this point. They can afford to flail around to see if something sticks. Reminds me of Rich kids who have ability to travel the world and find themselves before settling into a career
One problem is that the AI agent market is fiercely competitive. Why build when you can buy? For the foreseeable future there will be a number of competitive models on the "efficient frontier" and I don't think one vendor will pull ahead.
In that market you can build a model and spend a lot of money on it and at best get something that's on the same frontier as everybody else but just as likely end up with uncompetitive models like the ones they have now.
You might save a bit running your own models, doing your own inference, etc. Why not take advantage of "last mover advantage" and buy whatever is best when you need it and figure the odds are good that everybody else is going to buy more GPUs than they need and as a large customer you'll be able to buy in bulk at fire sale prices?
That makes sense in a way, but remember that Meta had previously seen some brief developer glory in the initial Llama release. Going the off-the-shelf route would essentially be giving up on being on the technology frontier in this area, and not monetizing their knowledge assets.
1 reply →
> I think most of the substantive criticism of Zuckerberg has been about burning funds.
The 2017 Rohingya massacre in Myanmar? They handed him the death toll. He filed it under growth.
>I think most of the substantive criticism of Zuckerberg has been about burning funds.
I'm not in the org myself I know some Meta SWEs tangentially. My understanding is that the biggest criticism is just the chaos of it all. Jumping constantly from one thing to another like headless chickens and accomplishing nothing.
It created an environment where it's kind of impossible to plan and progress your career.
While I mostly agree with your post, I do want to point out one thing:
> Or he could release a model trained largely by existing open weights models. Which without some huge breakthrough probably has no chance of surpassing them, so is pointless.
This seems to be categorically untrue. Composer 2.5 is a substantial improvement on its underlying Kimi base model.
If that is backed up by benchmarks then maybe they should imitate whatever Cursor did. What did they do?
They may eventually have to do that. Or they might be starting with an existing Llama model. Maybe I should have said "huge breakthrough or additional dataset".
We're getting closer to the bubble-bursting peak. Yet another tech corp that fired a ton of experienced people hoping to squeeze more $ for a quarter, finding out AI hype is not all it's cracked up to be. Loving that for them.
Second thought, I wonder if zuckburger has turned into the Cramer of tech. First, the metaverse fiasco, now another massive overinvestment into a dead-ish end?
Ahh the classic bootlicker's bias, this is how oppressed developers move and click their sad cursor hoping to keep their job while they train their AI replacement!
>"For people who are comfortable, that's great, they can contribute to this kind of great human survey. To people who are not, it is not an issue,"
Maybe they'd make faster progress if they worked in the Metaverse.
You always get more work done when you have no legs to worry about
why havent big tech employees formed a union?
That's a great question. Some form of union has already started in some companies, as far as I know. But not many employees have joined those unions. Probably because most employees have high income and don't really feel like they need collective bargaining power, compared to other low-income laborers.
Poor skills at working together and love of high salaries over work place autonomy.
Fun fact: https://news.ycombinator.com/item?id=23496533
Did he announce it in the metaverse? Bro is never right.
Want with one hand, shit in the other. Which fills first?
i bet he wants some calculative shit
Or some fuzzy yet inevitably reliable shit.
The modern trend is to think intelligence is generative “like compression” or “predicting next in sequence” rather than iteratively reducing uncertainty, like those fault tolerant humans.
Compression can be defined as reducing uncertainty. If you can predict the next sequence you can compress it to 0 bytes using arithmetic coding. Reliable prediction is what enables compression and it's the link between compression and AI that everyone is talking about.
No one ever in comp sci says artificial intelligence is "like compression", they correctly state that "artificial intelligence IS compression". It's absolutely known and accepted that artificial intelligence (defined as predicting outcomes with a measure of certainty and taking chosen actions towards goals using those predictions) has equivalence to compression in a very hard science way. The hardest part of artificial intelligence is compression and the remaining part, the choice of actions based on predictions is just a tree search to a goal.
3 replies →
emergence is a good philosophical concept but really bad if done in engineering that is why i never develop llm but hardness or use llm at most
i blame Wang!
Wang should be CEO. Zuck is the one at fault here.
[flagged]
[dead]
[flagged]
[flagged]
[dead]
[flagged]
[dead]
[dead]
[dead]
[flagged]
[dead]
[flagged]
Isn't he just trying to tank competitors prices? Try to burst the bubble in a race he is way behind?
[dead]
[flagged]
Although there is a difference in accrual earnings and cash flows. I was wrong with the number provided which although not as big as 1.5B of AI spend, is still a comparatively large number itself.
I have written an more in-depth comment if that interests ya (in a good faith discussion and please be kind to everybody)
also please don't call @kentonv an idiot and please read the HN guidelines[0]: Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.
[0]: https://news.ycombinator.com/newsguidelines.html
Would you be happier if I said "non-GAAP profitable"?
[flagged]
You are saying $1.5B of AI spend would not have shown up in either of these numbers?
Please do tell us how that works, I'm sure it's fascinating and would love to know more.
How does he get to decide what's "enough"? Reality will tell us, he can only place bets, whether it pans out isn't something that he has any say in.
Mark is really a bad leader with a mwah mwah vision. He is maybe correct in some things. But the execution is really really poor. Plus he does not have followers and believers. He only got money that can simulate followers to a certain extent
If it was still possible to get verbatim results from Google then I believe "mwah mwah vision" would have been an authentic Googlewhack pointing at this comment thread.
Nice thought, but I think strict googlewhacking frowned upon quote usage.
Actually, agents have been developing incredibly fast. It is just that Zuckerberg has some unrealistic expectations; the man is a lunatic.
Over the past six months or so, OpenAI's internal team has completely shifted from being heavy ChatGPT users to using Codex. Once you start using an agent like Codex, it is very hard to go back.This shift is truly transformative.
I am also aware that some of the consumer agent products on the market are growing very rapidly, such as Manus and GenSpark. Not to mention Claude Code and Codex.