Comment by jrflo
5 days ago
Still satisfied with my switch to codex/chatgpt. I couldn't imagine switching away from claude code when it first launch but with the drastically more generous usage on codex for the same subscription tier I just can't justify it.
My experience is that the GPT-family of models are very smart and figure out bugs, edge cases a bit better, but it produces code that is much less mergable – if you review the code, it introduces a lot more useless/inappropriate heavy abstractions and wrapper functions, compared to the Claude-family models which introduces the right amount of straightforward human-style code.
I can recognize so much of the GPT/Codex generated code long after it gets merged (not by me).
Additionally, the time spent on every agent turn on GPT 5.5 is much longer compared to Claude Opus 4.8, which means iterating on the code takes a lot more patience, and there's a lot more nitpicks to pick when actually using GPT 5.5 to do software engineering.
Feels like GPT-style models are more geared on doing one-shot software vibing (and handling the vibe coded mixture) compared to Claude's focus on actual software maintenance. I got a GPT Pro sub for free and wanted to cancel my Claude subscription so much, but I still keep reaching Claude models a lot more. Frustrating.
"5. DON'T FUCKING OVERENGINEER! WRITE THE SIMPLEST CODE THAT CAN POSSIBLY WORK! NO NESTED LAYERS OF ABSTRACTION! NO UNNECESSARY CLASSES OR METHODS! NO DESIGN PATTERNS UNLESS THEY ARE ABSOLUTELY NECESSARY! NO MAGIC! NO SHENANIGANS! JUST THE DAMN CODE THAT GETS THE JOB DONE IN THE MOST STRAIGHTFORWARD WAY POSSIBLE! THE FIRST PRIORITY IS TO WRITE CODE THAT IS EASY TO READ AND UNDERSTAND AND READ!!!"
this is the line I keep in Agents.md that helps me prevent Codex from playing smart
The urge to put capitalized, repetitive, borderline abusive instructions should be studied. I haven't read many academic papers looking at the frustrations around repetitive patterns.
12 replies →
I have a theory that swearing actually results is less comprehension of instructions by the model due to lack of training data over more conventional MUST.
We were reviewing reports of situations where the models failed to follow directions and there was a common thread of some where when the operator got the model to acknowledge the rule breach, it quoted back something that included swearing.
I don’t have the data to truely look into it, but I did give the instruction to my engineers to avoid it as a “might be a problem”.
11 replies →
You haven't really lived until you've had to type this whole thing, aware of the fact that the all-caps doesn't change much, but they stay because the rage has to go somewhere
Bonus points if you find yourself actually saying it out loud while typing it.
I have used the word "shenanigans" way more in a couple of years of agentic coding than in 30 years of writing code with humans.
Will save you some tokens: „write code like Linus Torvalds” - model should have all his swearing included in training data.
I have found many mode of failures with Opus during some task related to writing letters (not legal), and I actually put it into the memory and it works more or less for these specific tasks. For example when I want it to draft something, it always ends up being so flat, yet when it explains them to me, it is usually really great but not when I am telling it to put it in the draft. Adding these to memories with the help of Opus ended up resulting in a much better experience. There are still some blind spots but I also figured out how to make it give me the charitable version, without less protection, so I do not have to now go back and forth it.
I noticed that when trying to use Codex and compared to Opus. So many layers of simple functions added by Codex. I need to try this out in my Agents.md.
Curious : why would you say no design patterns?
1 reply →
i actually think this is too tame. it really has to be stuff youd mever say to a real person.
1 reply →
It might be a salient point but I didn't read it as it was yelling at me.
you forgot to sign it with Donald J Trump
1 reply →
I'm not sure if i do something differently but i have the exact opposite experience with these models. Claude always feels like it's generating way too overdesigned and hard to understand code with the vibe oriented feel while codex is cleaner and more "task at hand" and easier to work with.
Agreed
I echo your observations. I expect you will enjoy deepseek-v4-pro for writing code. Much closer to that Opus experience, and very cost-effective too. With 5.5 as a reviewer and specialist, all bases are covered.
Have you tried iterating on style feedback in AGENTS.md? I've been reasonably successful using this to get it to output code in a terse, non-defensive style that matches my hand-written code.
GPT-5.5 did a significantly worse job than Qwen-3.7-Max on a job today (some devops tasks I wanted to create some reusable scripts for). Kind of disappointing.
I've also seen Qwen 3.6 beat GPT 5.5 a couple of times. The ball is definitely in OpenAI's court now. Qwen is not going to fare so well against Fable, from what I've seen so far.
1 reply →
This is my experience as well. I have defined a CLAUDE.md rule to ask codex to automatically code review, and I tell it that the reviewer is very picky and to only implement what it considers valuable feedback. I hope they don't converge over time, currently, in combination they works really well.
i had this same complaint but no offense to you it turned out i was just not using the models right.
ai llm are doing what i tell them to.
if you’re building something meaningful (in my case a platform used by many people across many companies) you want to ensure you
1. have actual systems engineering and architecture in mind that you want the models to
2. implement based on what you tell it to do
when i was just telling the models what i want done without doing due diligence it would go and do some moronic implementation that was awful. mid input = mid output
these days i just maintain specifications documents and the AI follows everything i tell it to in that document. so when i tell it to dos one thing, the result is made following those architecture specs.
i have code that is single resp, modular, easy to extend and test.
i would ballpark 95% of the time i get what i asked for.
sometimes it tries to be clever in cases that weren’t covered in my arch specs. in those 5% of cases i go and update my specs.
source: used billions of tokens worth to build something actually in production across both mobile platforms and web, deployed on my own cloud infra. i use codex mainly. some claude.
I noticed too, that whatever they offer in the chat, for free, is smarter, as in no more bs. I use claude code and I want to try codex too but I don't need two subscriptions. I did try codex for some planning and it was really good. Thanks for giving me an insight into how it generates code.
Codex IME is just smarter, I think it shows given both anecdotes but also how OpenAI has always been at the front of programming competitions and math problems.
But Claude models seem to be better at long term problems or more ambiguous problems.
I'm curious as to what the primary benefit here. Are there secret improvements in training? There hasn't been much in fundamental model architecture, I don't think. What about harnesses? I wonder what's pushing the AI. It seems like harnesses is the main thing pushing AI ever since CoT.
I find that OpenAI's agentic tools and models are better for building human-maintainable software. Meanwhile, Anthropic seems to be cosplaying Apple while missing out on all the exceptional engineering required to create something that polished. Their admission of predominately using Claude with little human oversight and their stealth mode is an indictment of a poor engineering culture, from what I can surmise.
Serious question: what is the secret to getting Codex to write decent code? I am on Windows. Maybe that is the issue, but I can't seem to get Codex to function anywhere near the level that I was previously able to get with even Claude Sonnet. Does Codex just not work well with Windows yet?
4 replies →
I've had the exact opposite experience. For various reasons, I've had to move from Claude to Codex and the rate at which it burns tokens for the same output I would get from Claude is ridiculous. I'm probably burning tokens at a rate that is at least twice as much as I was when using Opus 4.5 for coding tasks and still finding that just manually coding is easier than trying to get Codex to write functional code.
How smart a model is varies hour over hour, tracked over here: https://aistupidlevel.info/
I guess enjoy it while it lasts? OpenAI won't be able to subsidize that forever either.
Agreed. I think the Chinese labs are proving that OpenAI and Anthropic don't have a moat in almost every aspect, especially pricing. I also think people are getting annoyed with the constant lift and shift. I've seen more folks drop Claude Code and Codex, specifically, because of the lock-in it provides the providers. I'm curious to see how people standardize on tooling adjacent and if Anthropic, Google or OAI move to block utilization akin to the games Anthropic has been playing as of late.
I think the end game is routed model usage and SLMs. I think Apple is going to prove this in the consumer space pretty handily and I'm curious how the Android ecosystem responds since the hardware is considerably lacking in model performance. I think Apple has a huge opportunity here, as much as I don't like their current ecosystem of walled garden. They did position themselves very well with ARM and custom chips for their hardware. Hopefully the broader ecosystem of ARM and Linux are able to make some headway and we see a more formalized, and broadly accepted, architecture to capitalize on.
is there an alternative to codex that “just works”? by just works i mean i can install as an app in 1 minute, and i get web search, skills, mcp servers, etc? Bonus points if it can control my chrome tabs like codex can, and if it offers remote control from my iPhone (chatgpt app) so i can kick off tasks while i’m out for a walk. Even more bonus points if i can, with 1 button click, share my chats or share the results of a session as a “site” (vercel style).
I’m sure you could put something similar together with a bunch of duct tape and 2 weeks of effort, but it won’t work nearly as nicely nor out of the box. so…what am i missing?
1 reply →
Big companies are not doing OpenRouter.
My company has an agreement with the big providers and while i'm pretty sure they think about how to get budget back, its an competitive advantage and normal people will not learn different model behaviours.
At least for now.
1 reply →
I see exactly opposite . Chinese models fails under any complex scenarios, while us labs raise the price , that's a sign of confidence.
5 replies →
What lock in does codex have? I'm using it it pi harness specifically because it doesn't have much in the way of lock in.
I don't think anyone has a firm grasp on actual inference costs -- including the research and training that has gone into those models. We've got near-frontier capabilities from open source models from China at pennies on the dollar compared to US big tech rollouts. OpenAI and Anthropic are heavily subsidizing their inference -- no wait, they are charging the most they can get away with before going public. Where is the truth?
> I don't think anyone has a firm grasp on actual inference costs.
There are huge numbers of users (myself included) that do have an exact idea of what inference costs are - on open models. We can buy tokens from 3rd parties that have no motivation to subsidize our use. That's to say, there's a fair marketplace[1] and we're hanging out there.
If you want to say "I don't think anyone has a firm grasp on actual inference costs on these proprietary/closed models", then I could agree with that.
[1]: https://openrouter.ai/rankings#leaderboard
Both can be true. They can be charging what the market will bear, and still be charging less than their costs of running it.
7 replies →
> OpenAI and Anthropic are heavily subsidizing their inference -- no wait, they are charging the most they can get away with before going public. Where is the truth?
Both. They are charging the most they can get away with and that amount is still heavily subsidized by VC capital.
> I don't think anyone has a firm grasp on actual inference costs -- including the research and training that has gone into those models
We know roughly how much these companies spend and what their revenues are. Based on that, they'd have to more than double revenue (without spending more money) just to stay even, and that's not good enough given how deep in the hole they are.
> OpenAI and Anthropic are heavily subsidizing their inference -- no wait, they are charging the most they can get away with before going public. Where is the truth?
Both are true. I mean, I'd be willing to spend a bit more than I do now, but not more than double, and neither are most companies. The company I work for is currently investigating how to reduce LLM spend, not looking to spend more.
We have a firm grasp on actual inference costs from the various open weights model providers on OpenRouter. They don't have the money to subsidize inference and it's quite a competitive market, so the prices are representative of the costs.
We pay by token at work. I just finished one session with Opus that was 4000 dollars. In about three days.
Now that 200USD subscription starts to feel cheap...
12 replies →
[flagged]
2 replies →
I'm planning on switching from the $20/month to the $100/month plan.
It's worth it, and I can afford it, but I am not really the right type of user for token-based usage. It's all for personal and free work.
Just a personal anecdote but I have not hit any more thresholds or limits since switching to the MAX plan and so far, it's been worth it. But I do wonder how long even this will last...
8 replies →
I have the $100 plan and had almost never run out of credits until I started using the ultracode / workstreams feature w/Opus 4.8..at which point I managed to blow the full 6 hour allocation in like 20 minutes, or so. In fairness, it did some amazing things with the extracted information, but it also strongly suggested that I'd need the $200 subscription *plus* a budget for extra usage.
Instead pay for 3 Chinese models. No max out ever then. I pay for kimi, DeepSeek and Claude. Whenever Claude decides it's over, I can safely continue on very cheap plans.
My bet is they'll keep subsidizing for a considerable period of time, at least 1-2 decades more.
Most AI companies are just testing the waters with paid tiers right now, their greatest fear with increased pricing is folks reverting back to wikipedia, stack-overflow and other public domain organic activity buzzing back to life; that will kill any RoI potential in LLMs forever. They're playing the wait game instead, observing how the digital sphere reacts to every little increase in price.
If that weren't the case, they'd be pricing at lucrative premiums already and even gotten away in short-term considering the increased dependency in the enterprise world. But that'd be like killing for the golden egg too soon and losing all long-term potential.
Once the folks are so addicted to LLMs that even writing a hello world program sounds like a nightmare and coming up with an article draft feels like reinventing Egyptian glyphs, that's when the real pricing hammer will come.
Anthropic and OpenAI won't be around in 1-2 decades if this is their long term plan. People are not going to revert, but go elsewhere. China is proving that it can be done cheaper.
1 decade = 10 years ...
Oh for sure. I've been hopping around from provider to provider for the last few years just depending on who has the most capable / subsidized plans at the moment. I definitely expect there will be a squeeze on subscription costs all around the industry post IPO.
A few weeks ago they massively cut usage on free tier.
Nothing is subsidized. Subscriptions are profitable for both Anthropic and OpenAI.
Anthropic wanting to switch billing to API rates is them just wanting to generate more profit.
> Nothing is subsidized. Subscriptions are profitable for both Anthropic and OpenAI.
Even if subscriptions are locally profitable (i. e., the cost of the subscription covers the cost of inference), they're still subsidized because they don't cover training and running the company; otherwise, these companies would be profitable.
2 replies →
"Nothing is subsidized"
So they are profitable?
I think you are mismatching accounting terms.
You can't say the 'subscriptions' are profitable without accounting for the cost of making the model that is the source of the subscription.
They are heavily subsidized by the shareholders. Investing, running at a loss, with hope of some future profitability.
1 reply →
"Nothing is subsidized" is a wild take. They might be making money on some users, perhaps even most users, but certainly not all. Also, "subsidized" doesn't just mean on compute.
That's interesting. Do you have anything to back that claim up?
3 replies →
100% I constantly get errors and timeouts on single responses in Claude, and certainly hit limits all the time. Codex rarely. In fact, I bought a second $200 Codex plan because the quotas seemed fair and I didnt have constant issues. Claude is so great at a lot of things, but unfortunately Anthropic beats you away with a stick every chance they get.
I've only ever had the $20 month claude plan but last night took the time to setup opencode + openrouter paying for deepseek + glm. Previous experience, while extremely awkward, I'd hit my limit within one or two chat replies and it'd take me like 4 limit cycles to complete my task. Now I'm able to complete an equivalent task entire task for less than $2 in two cycles (ask -> revise).
I'm doing basic web development here utilizing animejs. Nothing too complicated (mostly saving time doing the scaffolding, still write the bulk of animations manually).
Truly believe that American companies are going to get completely curb stomped by China due to greed, ineptitude, and violating the social contract.
I've switched from OpenRouter to using Deepseek directly from their platform since OpenRouter providers were pretty flaky and inconsistent.
Deepseek V4 Flash is suprisingly capable and insanely cheap. It takes so much to get the session cost to get to $0.01.
Nice, will do this this weekend. Been very impressed with deepseek. Did like 8 hours worth of work after that post and it costs less than $3.
The openrouter provider flakiness with deepseek was infuriating, but I’m happy in hindsight because direct deepseek has been very pleasant. Shocked by how low spend is.
> and violating the social contract.
I agree with you on pricing, but what do you mean by this?
Sure, modern American corporations care more about hoarding wealth rather than helping build up US society. Once neoliberalism became the mainstay economic position of the US income inequality has skyrocketed, healthcare costs have increased, childcare is more expensive than university, housing has become both unaffordable + unobtainable. By simply existing costs have increased while life becomes unstable.
Why aren't corporations doing more to help workers with childcare? Why aren't they doing more profit sharing with workers? Why aren't they encouraging unions or sectorial bargaining? Why isn't the government mandating any of this?
Americans very rarely benefit when US corporations do well. That needs to change. No one benefits if Meta continues making billions in profit every quarter while society suffers from isolation, depression, suicide, and scams from their services. Americans don't benefit if health insurance companies are making massive profits while they can't afford deductibles.
Our society has been setup to simply extract wealth in all facets of life. That's a sick society and it needs to change.
I'm not saying China does this better, in fact China has some of the worse worker rights out of all the industrialized countries; but at least American consumers would benefit from cheaper higher quality Chinese goods. The world would likely benefit too if America got off the cold war hype train that did nothing to benefit humanity outside of those making weapon systems.
2 replies →
I have been using both codex and Claude in my day to day, trying to not get to attached to one. I want to be able to work with any provider in case one of them does something bad.
I feel like Codex made a big push to run everything on your laptop. With Claude, I get 4 cpu's, a fair amount of ram and 30gb for every one of my dumb ideas for free in the cloud containers. Codex used to be similar, but last time I tried it just kept pushing me to run it locally on my laptop, which I really did not want to do with 20 requests going at once. That's the main advantage for me at the moment.
What runs in cloud containers? The dev servers, builds, etc.? I tried to quickly glance at the Claude website and it doesn't mention cloud containers on their pricing page.
The dev environment runs in the cloud. Like devcontainers if you’re familiar with that, except the IDE is just the Claude app.
Having said that, I found the cloud dev environments slow to the point where I wasn’t sure if it had frozen, so I never looked back.
"cloud containers" do you mean Claude Code on the web? Codex also has similar Codex cloud.
Yes, correct, they both have the same capabilities, however it felt like codex was pushing me harder to use my local desktop in an annoying way, while claude code was happy to spin up a bunch of dev containers for me in the cloud.
I've found Codex to be the better subscription for OpenClaw, because the limits are indeed very generous. However, I've found more and more that Claude Routines/Scheduled agents can replace all the tasks I use OpenClaw for, so I've been slowly switching over to Claude Code. Aside from OpenClaw, I don't find a lot of value in Codex as a harness on it's own.
I have trouble justifying gpt after that gross stuff with the war department.
Though the day is coming when there’s no distinguishing, I’m sure.
Right now there are Anthropic engineers deployed in the NSA to help them use their cyber models. The NSA is part of the department of war.
pedantically, the defense department.
"War department" is the older name, not "Defense department".
Also, is it really a defense department when you're starting wars of aggression every 15 years or so?
6 replies →
I do slightly prefer 5.5 for complex work but Claude quota usage has gotten infinitely better since the dark days a few months back - has gone from being infuriating to something I pretty much don’t have to worry about with it as a daily driver. (In fact, hitting GPT weekly quotas is more annoying now). Understand if people are still scarred by the issues + poor comms around them, though.
That's good to hear. It was legitimately unusable back when 4.7 was released, so I had no choice at the time. I'm sure I'll ping pong back again at some point.
Do you use a token service like open router or just subscribe to / unsubscribe from various models sequentially?
I just subscribe/unsubscribe to the providers each month. I'll definitely check out open router though, I always assumed that subscriptions were heavily subsidized by the providers especially if you're on the top end of users but maybe I should go to a usage-based plan.
Wait till you kick the tires of Qwen Coder.