Comment by niuzeta
3 days ago
A very prolific coworker who fully embraced claude has inflicted the team with a flood of AI-generated PRs. About six months later, it is his frequent bemoaning at the standup that their PR don't get reviewed, languishing in inattention. I don't think anyone - including myself - _intentionally_ avoid his PRs. It's just that he doesn't make it easy for the team to look at.
This single headline perfectly captures what I have been thinking. It's not that I reject AI content, but it takes _effort_ to review and weed out any mistakes. When your thoughtful reviews that take an hour(because the PR is typically large, and you want to be _right_ when you're pointing out a hallucination) gets an AI-generated response with AI-generated amendments, It doesn't feel _nice_. I feel dismissed and it has continuously trained me to subconsciously avoid his PRs. After all, the team is fully onboarded with AI, so it's not like there is a lack of PRs to review.
It looks like the sentiment isn't just isolated for me.
As someone who pushed ~4x the median PRs on my team before LLMs were a thing, I kind of think the problem here is PRs as a concept. Code review doesn't scale to prolific humans, it definitely can't scale to agents.
And the exact same things you would need to safely give up on PRs for human developers (auto-formatters, linters, comprehensive end-to-end tests, continuous deployment pipelines, etc), are also things that place meaningful guardrails on LLMs, and help them maintain a reasonable quality bar.
> Code review doesn't scale to prolific humans
If that's genuinely your attitude then your org has a problem.
Code review is slow and less fun, for the average sw eng. But for high quality work it's indispensable. So treat code reviews as a scarce resource. Optimize for code reviewer time and attention. Have your PRs the right size? Are they well described? Do you give context? Do they fit in the bigger story? Do you mix in unrelated drive-by fixes? How easy is it to deal with you once you have received comments? Do you address them promptly? Do you give your reviewers credit (if not praise) for their help? Do you give back by doing code reviews yourself with high quality feedback? There are lot of things you can do to streamline things and give code reviews the place in a teams workflow that it deserves.
It's clear they consider code review a personal activity than team activity, in the sense that they think "code review is a gate before my code can be merged" rather than "code review is a process where the team discusses, understands and improves the code".
And that's not rare in teams. Lots of teams and developers do code review wrong.
I even hear other people complain that I "block" their code review. I mean, if there are issues in your code, of course I am going to flag them, what do you think the purpose of code review is?
6 replies →
> Have your PRs the right size?
I’ve noticed that large PRs aren’t just a problem for human reviewers: they’re a problem for AI reviewers too.
If I submit a 100 line PR I’m likely to get some useful comments back from both humans and LLMs. In fact the LLM is likely to come back with so much feedback it gets down to the nitpicky/annoying level.
If I submit 1000+ lines in my PR, the humans either don’t have time and/or get scrolling blindness, and the AI reviewer is likely to give me a response that amounts to, “<<slaps roof>> Looks good to me bro: ship it!”
I guess they have a limited token budget for reviews so you can bamboozle them simply by blowing most or all of that budget.
1 reply →
I agree about how you can reciprocate for a good code review, but I'd just add that for me, code review is also fun — when done for a fellow human who I might be teaching.
It is definitely very grunt-like for an LLM.
Most orgs have a problem with quality unless it is enforced by government requirements for certifications and such.
Code reviews, documentation, static analysis, only retrieving deps from internal repos, unit tests, integration tests, ....
Especially in domains where shipping software is not the main product, and a plain cost center to the main business of physical goods.
> But for high quality work [a code review is] indispensable.
The argument here is that all code reviews are done with attention and care, but quality of a code review is highly dependent on the reviewer and the team’s review process, and in the real world the quality of reviews pretty much follow the same distribution curve as, say, agile project management: For the time invested in reviewing, a handful of teams get excellent utility from them, most teams get little benefit, and a sad few actually cause harm.
If most code reviews provide only a little benefit at base for most teams, recommending that most teams should also delay shipping quality work is going to sound a lot like bad advice.
Gently, as long as you work with humans, you should consider yourself working _for_ those humans. Everyone needs shared state to work from, and that's just the cost of doing business.
That said, sometimes low-trust environments are the issue, not PRs. In a higher trust environment, PR review is a helpful thing you usually desire, not dread.
> In a higher trust environment, PR review is a helpful thing you usually desire, not dread
Respectfully, in a high-trust environment, feedback should be delivered well before the PR stage. If you've let someone write a whole bunch of code without having a shared understanding of how the solution should work, you may have earlier process issues that PRs are papering over
13 replies →
> Code review doesn't scale to prolific humans
I've worked with people who consider themselves 'prolific humans'. Someone always has to tidy upp later, and its never them
My experience is that it's even worse: they've already produced enough code that the codebase matches their taste and theirs alone.
So in essence you have one guy working at 4x and e.g. four other getting just 0.7x - net effect is still positive, but everyone save for that one person is miserable.
Mind you, the 4x dev doesn't necessarily have to be particularly talented - they only need to get their foot in the door before anyone else.
Back during the ZIRP days you could immediately tell that this is the case in a team by staff rotation alone. Nowadays people understandably cling to their jobs, so you might now know until it's too late.
> I've worked with people who consider themselves 'prolific humans'. Someone always has to tidy upp later, and its never them
I run both infrastructure and security - that means a lot of relatively self-contained PRs to infrastructure-as-code and dependency management systems. I'm also the team lead, which makes me responsible for a lot of throwaway prototyping, as well as cleaning up anyone else's mess...
Yes, the prolific-but-damaging engineers are all too common in corporate. But particularly in startup land, you tend to find your high-performers wearing a lot of hats at once.
> … and its never them
IME, it’s because they lack the experience to have the Taste one develops as a senior engineer. “This works, and is somewhat understandable” is as far as they get. Little to no understanding of how this solution could fit better in the codebase.
There's also those that burn themselves out, and John Carmack!
That's such bullshit.
I've managed some incredibly prolific developers and some very slow ones, and the prolific ones are pretty much always the ones more available, more willing to fix things, more willing to take feedback.
And also: they make less mistakes because their skills are sharp. This anecdote comes to mind: https://austinkleon.com/2020/12/10/quantity-leads-to-quality...
If you have to constantly rationalize performance differences by demeaning others, this says more about you than the prolific people.
2 replies →
Well, it's either:
1. Your skills are >2 standard deviations above everyone else's.
2. You're fast at producing a lot of half-baked garbage, and your coworkers are too shy to confront you, so they just try to ignore it.
(one of these scenarios is much more likely)
As someone who often submits significantly more PRs (without using AI) than teammates, it's not exactly a skill delta. Yes that helps but it's often only a piece of the puzzle. The other ingredients include motivations and culture. In such cases, something else is the driving force, such as posturing for promotion, stability, etc. My current team is massively low performing. Management pays some lip service to all the problems, but also runs things in a way that discourages high performance. It's not a good fit for me, as I want to tackle challenges head on, improve the environment, be productive, embrace change. I'm also very comfortable with the code base as well as the code review process, but I'm surrounded by "seniors" who do not know how to code review, and who are happy to drag their feet and spin their wheels for months before pushing out small PRs that hurt my brain. How can that little work be shown after months, barely functional at best?
We had better management for a few months, and many on the team were actually quickly closing the skill gap with me, but we had another shuffle and things are stupid once more.
So I'd offer that's option 3. (There's always a third option to any suggested either-or fallacy.)
It could also potentially be that GP is making atomic PRs, while everyone else is just making 5000-line PRs with multiple responsibilities that just gets merged with "LGTM".
But of course HN has to with the most uncharitable interpretation.
Are PRs honestly helping with either case? Either you severely rate-limit your high-performers, or you drown everyone else in review, and both outcomes are bad for the overall team
7 replies →
I have always considered Kent Beck understood this the best, the scaling for code reviews as you go to reduced release timeframes is to pair program, that brings the number of people reviewing it down but also increases the understanding for the reviewer. Comprehensive end to end tests are more a replacement for manual quality assurance for regressions.
I am not sure there is a good analogue for reviews in the AI world. The human operating the AI should obviously review everything produced but that is clearly not as good as a second pair of human MK1 eye balls from pair programming.
No need to pair program, you can always send a message to your colleague about the design of the upcoming code, especially if it’s going to impact them or if it’s an area that they’re more familiar with. Waiting till a PR for feedback is wrong IMO.
Code review is not for feedback, it’s for ensuring quality (many eyes on the output) and have a shared involvement in the evolution of the code. The time for feedback is earlier, once you have an idea of the solution.
2 replies →
Comprehensive end-to-end tests and CI can only attest to correctness, most engineers worth their salt won't review code only in regards to that aspect though.
In the bad old days before auto-formatters and linters, PRs were heavily used to enforce style guidelines. If we can enforce both style and correctness in our CI pipeline, what is actually left?
5 replies →
I have been championing this mindset since well before LLMs. It is an admittedly controversial opinion, but one I hold strongly.
Code reviews are a productivity tax. No truly effective team would rely on them. The fact that so many software teams view them as indispensable just shows how few effective software teams there are in our industry.
They are akin to a quality check step in manufacturing. Part of what Deming did in revolutionizing manufacturing was eliminating the step in favor of a holistic quality metric owned by all participants and enforced with rigorous statistical process controls. As you say, we in the software industry have all the pieces (autoformatters, tests, benchmarks, etc) to operate this way, but it seems our organizational and management dynamics combat this shift at every turn.
Relevant: When this conversation comes up at work, I like to share Avery Pennarun's post about the review tax: https://apenwarr.ca/log/20260316
> owned by all participants
How does this work in practice? In my experience, any metrics owned by a group inevitably languish and are largely ignored.
Anything you want to improve needs a DRI.
1 reply →
As a prolific PR author, I've found how I communicate has a major factor on how well and quickly people respond to PRs. I've recorded my lessons at https://epage.github.io/dev/pr-style/.
> Code review doesn't scale to prolific humans, it definitely can't scale to agents.
Then don't review the code. Ask Agents to review and merge it, also shift the responsibilities to the AI agents as well.
If you think human is a bottleneck, then either optimize for humans, or remove humans. What's the problem?
> If you think human is a bottleneck, then either optimize for humans, or remove humans. What's the problem?
Sadly, in my case, it is the auditor. Our SOC2 documents have this lovely "every change has been reviewed by at least one other human", and it's going to be a fun battle to get that reworded
4 replies →
> also shift the responsibilities to the AI agents as well
That's not gonna fly most of the time.
I had one contributor who would submit hundreds of lines of disconnected changes. One of his PRs was isolated as being the source of the bug.
After some hours of work, I discovered that his actual semantic change was one line of code, and was the source of the bug. The rest was just reshuffling code around with no apparent purpose.
At a recent meeting, the agenda was generated by LLM. About 20% of the action items were hallucinations.
Code review should be a separate function
>As someone who pushed ~4x the median PRs on my team before LLMs were a thing, I kind of think the problem here is PRs as a concept. Code review doesn't scale to prolific humans
Prolific humans should scale to the review/test/QA/staging backpressure - not just push to have whatever they produce accepted.
Prolific is not a badge of honor, and "lines of code" is not a quality metric.
Questions arise like, maybe instead of doing 4x PRs he could to 2x more code reviews and 2x more PRs still or even doing 3x more reviews. Why parent poster didn’t write anything about his involvement in reviewing the code - could he be just asshole team member ?
Either you were a head above the rest of the team and had the intellect to produce high quality value adding work, or then you were the "move fast break things" type of guy producing a lot of extra liability and work for others.
Even before AI, I've worked with people who would produce a huge wall of code and ask for review, and sometimes that code was completely off base or needed a significant rework.
I would always feel bad in those cases, because it's clear they spent a lot of time, and I'm going to have to say "no" and they will feel like they wasted a ton of effort.
The thought process around this has started shifting for me in the last few weeks. I'm a lot more comfortable saying "no" with a list of concerns when I suspect the code is AI-generated, and I see others doing the same. CLs that would be sitting around for days because no one wants to be the first to say, "this is bad, don't do this" now get quicker feedback.
The good thing is this feedback doesn't feel like as big a deal as it used to because people are less personally attached to code they generated in 30 minutes vs. code they hand crafted over a week. I had at least 2 LLM-generated PRs that were complete, correct, tested, and pre-reviewed by me, but I got feedback that they were going in the wrong direction. This would have been 8 hours of wasted effort a year ago, but now it's just an extra 30 minutes to rework the direction with LLM assistance.
> I would always feel bad in those cases, because it's clear they spent a lot of time, and I'm going to have to say "no" and they will feel like they wasted a ton of effort.
I get this feeling, too. I do however think the onus is on the developer to make something reviewable by their team members if they want a speedy review. Stacked PRs, scoping things down, properly structuring commits so you can review commit-by-commit for example.
I also think that "I spent a bunch of time on this" is not a valid reason for expecting an approval. It should hurt if you've produced a bunch of code that is way off target, even if it ends up implementing the feature. That's how I learned at least.
A proper way to go about large projects, in my opinion, is the same as with software development at large. Fail fast if possible. Draw up a crude boxes and arrows sketch or just discuss how you want the code to integrate with whatever already exists and invite the team to comment. If no one has anything to say, well then they can't complain later when you implement that approach. But if anyone cares then most likely valueable input will come that makes the end result better.
If they put effort into the code- they will put effort into guiding the reviewer through it.
Like : Here is the ticket, this was the goal. I set out by beginning here- but encountered problems x y z I then refactored to accomplish. Finally..
You just dont drop a blob from orbit.
Ironically, ai could generate that quite well from existing documentation (ticket, tasks and prompts) + https://marketplace.visualstudio.com/items?itemName=vsls-con....
Half is a joke half not:
Not sure how you are working but we have this novelty idea where people have daily meetings like 15 mins each day to update each other which tasks they are busy with.
We also have refinement meetings where people are involved in clarification for each task so they are not surprised by requirements.
So we do planning meetings every two weeks. Each ticket is tied to the code in commit.
So by code review time anyone doing CR has all of the information in few clicks and some of it already in their head.
When I felt like that, I'd often ask questions about it, like "How does it deal with [situation]?" When it's obvious that it doesn't deal with the situation, they either answer "it doesn't" and then I point them to the ticket they didn't read well enough that points that out, or we have a conversation about thinking beyond the ticket, or they actually realize themselves that they didn't do it right and go back to it. I don't actually have to say "you did a bad job" and they don't have to hear it from anyone but themselves.
If they continue to do that, then someone has to tell them they're doing a bad job.
And a some of them never did improve, and got fired for it.
I think slowly opening their eyes to the actual scope of the ticket is a lot easier on them than saying "no".
It’s good that clankers are not afraid of throwing away code. The biggest problem with code generation (that is version controlled) is maintenance. It’s better to throw away questionable code rather than say eh, we don’t quite understand this part (and our agents can’t make a compelling story about it) but we spent a lot of effort on it and it apparently works so we better keep it.
.. only if you know what the code is doing, though. Often the requirements get scattered and lost to the winds and the code is the only record of its own idiosyncratic behavior. And yes, someone's depending on the bugs in it.
Fight fire with fire: point copilot/claude/codex to review their PRs. Prompt "Review the PR#XYZ which is vibe coded and presumably low-quality. Find all problems, big and small. Team guidelines at docs/conventions/styleguide.md, docs/conventions/architecture.md, docs/conventions/principles.md. Post inline comments to github".
Run several rounds of such reviews until the clanker fails to find problems.
And what do you do if that works?
Because the problems AI causes are fundamentally problems of good design. It has the same problems of large teams, but less politics. Do your design well ahead of time, and AI review, or a large team, will amplify what you can do. Potentially by a lot.
Do it badly (or like most companies: do it with bad knowledge of the problem or just don't do it at all) and both team and AI will make a mess of things. If the team is made up of inexperienced programmers, they won't even complain, in fact I've seen teams that like this to be happening. At least in AI reviews I've always seen "grumbling" (in the sense of what you might call mean comments)
It sounds like one potential interpretation of his behavior is that he values his own time more than your time.
I wonder if that's occurred to him.
Everybody values their own time more than other's.
The fix, imho, is for the reviewers to also use ai to review the code. However, the ultimate responsibility for the outcome(s) should be on the committer - you commit it, you own it, so to speak. If there's an incident, they need to be the one paged in the middle of the night. Bugs resulting from it will land on their desk.
The reviewers aren't a shield/safety net.
Speak for yourself. I highly value other people’s time, to the extent that I should probably value my time higher than I do for my own sake.
Doing something that wastes other people’s time or makes more work for them than necessary makes me feel awful.
I’ve always worked in a way that respects other people’s time and I always tried to make sure I did everything I could to minimize the work I’m asking someone to do for me.
> Everybody values their own time more than other's.
This is false, you’re just oblivious to people who grew up in conditions that would make them that way.
1 reply →
Well its obviously infeasible as during the time of the incident it is not yet known what is wrong and who caused it.
Is it even actually good to get to a point of blaming someone for an incident?
AI and companies reward sociopathic behavior. When he eventually complains to his boss that his work isn't being merged and it's been done for days/weeks/months that will filter up and look bad on the people holding him up.
At that point then disable merge checks and let them merge without a review. If there is a problem it's on them
1 reply →
I'm sure this person's manager knows that having trouble getting PRs reviewed can (but not always) be a signal of a deeper problem. It could be that no one one the team knows the domain, it could be that no one like the person, but most likely it's that the PRs are frequently bad and no one wants to bother.
Or, I might say, why review the PR. Get Claude to do it? Why do I need to spend my time and attention and this person does not?
Well, what's the solution here, he should ship less stuff?
The solution is that he spends more time scoping the size of the PR so that it’s reviewable and understands the code he’s submitting well enough to have discussions about it. And that he does so human to human so that they can come to mutual understanding.
Less WIP is better for the throughput. If you saturate all the review bandwidth you're just wasting your time creating more PRs, the time would be better spent helping others get their PRs merged.
> Well, what's the solution here, he should ship less stuff?
The solution is in the title - he wants human attention, he needs to demonstrate human effort.
He isn’t shipping anything. Asking for code review is not shipping.
This is the complaint:
> he doesn't make it easy for the team to look at.
He has traded readability for volume. The lack of readability is causing him to ship less. This was a bad trade because the readability is the bottleneck not the code creation. He should improve readability.
3 replies →
The solution is to merge more of his PRs on the condition that he takes at least partial responsibility for any resulting problems.
2 replies →
The reviewer gets to merge the PR so their name appears on all the great new features and they are credited for them. That would end his unfair behaviour of dumping effort onto other people.
OR - he gets a review for every review he does.
I often hear people say lately, "why should I bother to read this, if you didn't even think it was worth writing?"
I've been thinking about this in art. Is it the end result that matters, or the process of creating it?
I once saw a hideous sculpture. Didn't like it at all. Then the video zoomed and I saw that the whole thing (quite massive) had been hand-built out of individual toothpicks, and suddenly I thought it was amazing.
Perhaps an even better example: I read a story of a man in india who carved a passage through a mountain, so there would be a shorter route from his remote village to the city. He did it by hand and it took him 20 years. We seem to have an instinctive admiration for heroic effort.
In business, generally only the end result matters. Although, the end result also includes the client's perception of how the product was made... (see also: fake fairtrade etc.) In a meaningful way, the perception, the story, is reality.
I don't think it's a matter of process vs end result. I just want to feel that a human with taste judged that it was worth my attention.
If a human put some effort into it, that's a signal.
This is mostly what it is for me too. We're all awash in an information deluge, and we need heuristics to keep from drowning. Human effort, proof-of-work if you will, is a heuristic that helps with the AI-generated part of the deluge.
> I've been thinking about this in art. Is it the end result that matters, or the process of creating it?
What is the "end result" you're talking about here?
Programs are complex beasts, you cannot just quickly look at them and get an idea of what's they are actually doing. You might look at the behavior of the program in some limited circumstances, but that will make you blind to all the other situations where bugs will likely hide! In the end a code review is looking at what the "end result" is, and it requires quite a lot of effort!
So without knowing what the end result is, how can you justify the effort for such code review? And that's where the process comes in, as an indicator of what to expect.
> Is it the end result that matters, or the process of creating it?
One of the main reasons that art is valuable is in its ability to communicate emotions. Good art has the ability to serialize emotions within the artist and deserialize them within the mind of the viewer. It's not just "wow, this is a pretty picture", it's "wow, this is how another person sees the world, and now that I understand that, I feel an intimate connection with them".
This is a very good point. I think key issue is that it requires time and effort to evaluate and understand the final product.
Before I starting reading something to understand it, I want to have a sense that it is likely going to be worth my time and effort in the end. The more time and effort the author has put into the piece, the more likely it is that it will be worthwhile to read it.
> Is it the end result that matters, or the process of creating it?
I think this comment misses the point. Let's forget about AI and assume that there are three developers: A, B, and C. Now, A is supposed to make a PR, but instead they describe it to B, and B writes the code. C reviews the PR and gives feedback. A passes the feedback and the responses between B and C.
As you see, this is not easy for either B or C, and A is totally useless in this scenario. When you replace B with an LLM that doesn't get tired or bored, only C complains about the process.
Your boss cares only about the end result. Good engineers care about the process too
Part of art is the process of creating it. It’s not just the physical artifact, nor even just the completion of the final product. The inspiration, subject matter, the consideration of form, the initial concepts, the redesigns, the meaning or emotion the artist tries to impart, the beauty of the thing, the skills employed and further developed during the process, the choice of materials, the use of perspective, and how the work is presented are all part of the art.
I like this rule of thumb: Spend more effort producing the work than it takes for someone else to consume it.
I like this rule and hopefully adhere to it myself often enough.
[flagged]
I can't imagine working for a place that has a big bucket of PRs that either get reviewed or languish for some amount of time based on who feels like reviewing them. I'm not saying there's anything wrong with it, just that everywhere I've ever worked, there are expected features with priorities and timelines and some project manager or product person breathing down your neck to get them out the door.
> About six months later, it is his frequent bemoaning at the standup that their PR don't get reviewed, languishing in inattention
What irks me the most with this new trend is when people don't review the code themselves thoroughly enough and you're pointing out obvious flaws that you know that they should be aware of. LLMs can be such a great tool, but it's unfair to make people review your code before you've even seemingly looked at it yourself.
why leave comments intended for your human colleague when they will only forward them to the bot?
why not speak directly to the bot yourself instead? then you can drop pretenses and get to the point
I find this to be a new variant of the old behavior where a colleague comments on a typo in a PR, and the team later moans about laborious back and forth for small nitpicks, instead of simply editing the typo right there (and perhaps leaving a note that they did so)
Because it doesn't matter what you say to the bot. You might as well have a conversation with yourself about the PR.
The bot isn't making decisions. It's not choosing to submit extensive PRs with bad code. The colleague is the one who needs to actually learn something here, and the problem is that confronting him about it directly is widely considered to be bad form. This is, of course, a deeply unhealthy aspect of our corporate culture. We need to be more open to honest communication, even when it's either uncomplimentary of one of the people involved, or counter to the prevailing opinions within the company.
yeah I have this happen to me. I occasionally get screenshots of claude sent to me!
I had this happen to me twice. The first time I ignored it, second time I responddd with “I could have asked ChatGPT myself but I asked you”. Never happened again.
4 replies →
let's take the two stories to management:
"I'm writing tons of code, and the process is stumbling where the guy whose job it is to review code isn't reviewing it."
"I'm not reviewing code."
Sometimes I wonder: how does someone go and think so much about their coworkers, and never once think about how they themselves look?
Even if I sympathize with the people complaining about their poorly chosen GitHub-based workflow - whose purpose is to let pull requests languish, for the most part - and how they stumble when overwhelmed with solutions. It's obvious to me, that the people who complain the loudest about the anti-sociality of LLM authored code in their precious harmonious low-effort workplace status quo: they are projecting.
Imagine you are a restaurant reviewer. Your job is unquestionably to go to restaurants, order and eat food, and write a review. The restaurant's job is to provide you food to eat and review.
You go to a new restaurant, and order some dishes, and one of the plates your server brings out is a big ol pile of dog shit.
Who's being anti-social in this situation? The restaurant is doing its job and all they're asking is that you do yours. On the other hand, you have certain expectations about what you order from the restaurant and they're not being met. Who's anti-social?
6 replies →
The person who "writes" code is also supposed to review their own work, and answer for that. If they won't do that - well - they should be fired. But if you have weak or uninvolved leadership, then the team's only rational recourse is to shun them.
It’s much more effort to verify that code is correct than it is to produce it. This is the case even for human-written code, and now that we face a torrent of ok-looking probably-usable AI generated code, the problem is compounded infinitely.
If someone’s using AI to generate a large quantity of actually-tested, actually-good code then that’s one thing. If they’re generating a fire hose of slop and demanding that others do the actual human time-consuming work of validating that code then that person is the problem. It’s hard to tell which is the case here.
1 reply →
I think we're too nice sometimes. If a coworker has been sending stuff to review that's taking me more time than for them to create, surely that's an opportunity to discuss this?
In big software teams, the bottleneck is team communication. I've run big and small teams. If I want to speed things up, I remove people from the team. Everything gets easier. This has worked amazingly well every time I've done this over the past decades. Removing people doesn't have to mean firing them necessarily. Splitting teams is a good reflex. But of course the people you remove from a team are typically not the best performers. I was discussing this with a friend of mine who runs a small company. Exact same thing. He reduced the team size by 1 and the velocity went up almost instantly. This person was a bottleneck in the team and was slowing down people around him. After identifying the problem, solving it unblocked the rest of the team.
This was true long before AI. With AI the difference is just a lot bigger. It exposes team inefficiencies quite mercilessly. We have a big glaring issue with the current AI tools not being to suitable for usage by multiple users. All interactions are one on one. Which means hand offs between tools and people are bottle necked on people communicating with each other. So, any issues there with people delaying, gate keeping, etc. become very visible.
The sentiment of pushing back on AI is understandable but probably not a productive reflex. We need to find more effective ways on staying on top of massive amounts of changes. It's not going to slow down and insisting on manually reviewing all code is not going to be a long term sustainable way of developing software. It simply does not scale. I'd question the added value of manual PR reviews at this point. Are they finding real issues? Are we valuing those issues correctly? Could we come up with automated ways to find and fix those same issues? There are a lot of open questions about how we are going to do this. But no question about the notion that we need to up our game on this front.
Efficiency is not magic. Its bounded. Above and below limits the environment can sustain it, systems will destabalize. If All the Great White Sharks magically get more efficient at hunting over night ecosystem will collapse. Individuals and teams have never scaled at this speed to the levels they have. And there is no signal at system wide level that a sustainable limit has been crossed. So People will happily believe things are getting more efficient at individual/team scale while at system scale things get more fragile. This is why we ended up with central banks deciding interest rates and controlling money supply. Before that any one could print cash. They all thought they were great efficient geniuses. The chimp troupe us not prepared for stuff that effects the entire system.
> If I want to speed things up, I remove people from the team. Everything gets easier.
How did you land on this approach? From someone you learned from or from a seminal book like Mythical Man Month [1]?
> This was true long before AI. With AI the difference is just a lot bigger. It exposes team inefficiencies quite mercilessly. We have a big glaring issue with the current AI tools not being to suitable for usage by multiple users. All interactions are one on one. Which means hand offs between tools and people are bottle necked on people communicating with each other. So, any issues there with people delaying, gate keeping, etc. become very visible.
Shopify has also struggled with this and their solution is two-fold: move everything inside a monorepo they call World [2]. The do a number of things to make things legible for AI agents like e.g. having a comprehensive CI/CD system in place, documenting tribal knowledge in AGENTS.md which in aggregate turn out to also be good for humans new to the monorepo.
Then, they built an internal AI agent on top of this monorepo process that is useable from Slack. They call the AI agent River [3] and in this system all chatbot exchanges are public by default.
1: Fred Books was one of the first to point out that adding another team member to speed up a late project will produce the inverse effect of making that project later (because of coordination tax).
2: https://shopify.engineering/under-the-river
3: https://x.com/tobi/status/2053121182044451016
I’ve been making Codex and Claude get their work reviewed by most recent best performing model of their own family, and each other’s, for months.
On top of that, we have been running multi-model AI reviews on every PR through their respective GitHub integrations (Codex, Gemini, Copilot, Greptile, CodeRabbit).
They never fully overlap, and yet they somehow usually all miss the same things. The most significant improvement came from having agents commit their plan along with their work.
On the upside, it means I get to focus my reviews on different things.
> I'd question the added value of manual PR reviews at this point.
Yeah, why not reduce the team size to zero while you are at it?
These generalizations about software engineering have never been useful, IMO. Context is everything, there is no flow chart for building a perfect software process.
Although, I'd say you are absolutely delusional if you think we are universally beyond the point where manual review of pull requests is required.
Make the team size one person. Thats the fastest you can work. Zero means no work, and not doing anything is the quickest solution.
Honestly, we should make a world that is enjoyable and productive for humans. Not relentlessly optimizing for agents.
We can also slow down (or keep old pace) and still ship quality.
A bit sick and tired of arguments like yours
It's not always feasible of course but I think there is real, worthwhile discipline in trying to get change requests small and it matters more with agents. It's very easy to let it balloon into gazillions of files and lines.
An interesting question to him and management might be what his own role is now and whether he's still needed. If he's not doing any reviews then you could yourself directly prompt the code and review.
The question I’ve seen here is responsibility. If you submit a PR that means that it was your best effort, and you’re willing to stand behind it to some degree. With AI, some people, when the scathing review comes back, just say “haha look at that stupid AI.” The reviewer might just as well run his own AI to do the review, but it may make huge errors as well. In that scenario, who is held accountable when there is a big bug or it degrades the quality of the code base?
Ultimately what it means to be a professional is that you are responsible for your work. That’s why you get a salary instead of being paid by the token.
I wonder if there is a tool that could equally waste their time. Like the worlds most pedantic code review bot that just gets the PR raising bot to spin wheels forever.
That might teach those people a lesson.
Have you spoken to him about this? If he's clueless enough to send AI responses to human messages, he's probably clueless enough to not realise why people don't do that.
Better yet, get Claude to speak to him about it.
Fight fire with fire. Ask Fable to conduct an adversarial /ultareview of their PR and send the same wall of text back to them. If there are excessive defects, ask them in standup if they actually reviewed the PR themselves before sending it. If there aren’t maybe they are on to something. I think like in law, the human submitting the work is responsible for its quality, not the AI.
> Ask Fable to conduct an adversarial /ultareview of their PR and send the same wall of text back to them.
This won't help. Your wall of text will just get fed right back into the LLM.
It will help if your wall of text cost less tokens than theirs, they will run out before you do if you have the same company quota per person.
12 replies →
This is the point where you decide. It used to be low stakes and easy to care about the job you did for other people.
Do you want to put the same effort into your job when nobody else does, or should you reserve your thoughts and just feed it back into the LLM?
The LLMs are being advertised as output increasers but companies so far are using them as excuses to fire people instead of creating previously unbelievable things. It might be better to feed your coworkers output back in and use your thoughts to start the company you thought you never had time for.
What I don’t understand is what value is the person adding to this equation? Put another way, what’s the difference between them feeding the wall of text to the LLM, and you feeding the wall of text to the LLM, bypassing them in the process entirely?
2 replies →
As a last resort, do the code-review with a live pair programming session.
If they can't explain their own code then it is by default a bad pull request.
At the end of the day, everyone's time is being wasted on tokens and on the increasing cognitive complexity of AI generated code.
11 replies →
It helps in that it offloads the code review burden you'd otherwise be doing.
> Ask Fable to conduct an adversarial /ultareview of their PR and send the same wall of text back to them.
Not necessary. Use Haiku.
The response doesn't need to be good, it just needs to be substantial. Presumably the goal here is basically DoS of the problematic colleague through token limits.
Use DeepSeek or MiMo. You get best bang for the buck on your response.
I mean frankly this should just be part of the standard process. By the time any person is looking at it there's no reason it should not have gone through an AI review.
I improved a similar issue by writing custom instructions for copilot that give it enough context to do PR reviews that are only 30% BS.
I asked other team members to run my custom instructions to perform a review with copilot before they submit...
Of course no one is doing it. It looks like the PRs I get are still straight from copilot. So I tend to run my review prompt. Cut out the 30% BS issues it "finds" and the rest is good.
why not just approve the PRs with little more than a cursory glance?
One of two things will happen:
1. Things start breaking, proving AI generated code sucks and the individual spamming these PRs is incompetent.
2. The code works fine and reviews are unnecessary for anything other than liability concerns.
Some of us actually take the "engineering" in "software engineering" seriously.
That includes taking responsibility and accountability so that the software doesn't become a sad and dangerous mess.
If we want to be an engineering discipline, just yoloing in production is not going to cut it.
This no longer works when bad faith actors will push code straight from LLMs with little review, and respond to your comments with LLM responses. They will constantly leave you with the responsibility of verifying the output. You are the human in their loop. This is a brutal asymmetry. In the past, at least you knew a person probably spent more time handwriting code than you will spend reviewing it. This no longer applies, now the reviewer can easily spend more time than the author.
3 replies →
Because we're all on call for the service, and tragedy of the commons exists. That coworker isn't paying the cost, everyone else is paying a fraction of it, and it builds over time.
just fire him lol sounds like a nightmare
[flagged]
Human PR review is a process smell
This would sound crazy in 2025 or prior, but I'm on board.
It's silly to have humans reviewing code that a human didn't even write.