391–400 of 1001 posts
Earlier quoted context omitted.
Dumb question. Can these benchmarks be trusted when the model performance tends to vary depending on the hours and load on OpenAI’s servers? How do I know I’m not getting a severe penalty for chatting at the wrong time. Or even, are the models best after launch then slowly eroded away at to more economical settings after the hype wears off?
We don't vary our model quality with time of day or load (beyond negligible non-determinism). It's the same weights all day long with no quantization or other gimmicks. They can get slower under heavy load, though. (I'm from OpenAI.)
PS - I appreciate you coming here and commenting!
Impressive results, but I keep coming back to a question: are there modes of thinking that fundamentally require something other than what current LLM architectures do? Take critical thinking — genuinely questioning your own assumptions, noticing when a framing is wrong, deciding that the obvious approach to a problem is a dead end. Or creativity — not recombination of known patterns, but the kind of leap where you r…
Have you tried actually prompting this? It works.
They can give you lots of creative options about how to redefine a problem space, with potential pros and cons of different approaches, and then you can further prompt to investigate them more deeply, combine aspects, etc.
So many of the higher-level things people assume LLM's can't do, they can. But they don't do them "by default" because when someone asks for the solution to a particular problem, they're trained to by default just solve the problem the way it's presented. But you can just ask it to behave differently and it will.
If you want it to think critically and question all your assumptions, just ask it to. It will. What it can't do is read your mind about what type of response you're looking for. You have to prompt it. And if you want it to be super creative, you have to explicitly guide it in the creative direction you want.
Earlier quoted context omitted.
> These feel like they involve something beyond "predict the next token really well, with a reasoning trace." I don't think there's anything you can't do by "predicting the next token really well". It's an extremely powerful and extremely general mechanism. Saying there must be "something beyond that" is a bit like saying physical atoms can't be enough to implement thought and there must be something beyond the physi…
> Besides, what is the human brain if not a machine that generates "tokens" that the body propagates through nerves to produce physical actions? Ah yes, the brain is as simple as predicting the next token, you just cracked what neuroscientists couldn't for years.
> We build Claude with Claude. Our engineers write code with Claude Code every day well that explains quite a bit
Also explains why Claude Code is a React app outputting to a Terminal. (Seriously.)
Memory comparison of AI coding CLIs (single session, idle):
| Tool | Footprint | Peak | Language |
|-------------|-----------|--------|---------------|
| Codex | 15 MB | 15 MB | Rust |
| OpenCode | 130 MB | 130 MB | Go |
| Claude Code | 360 MB | 746 MB | Node.js/React |
vmmap shows Claude Code reserves 32.8 GB virtual memory just for the V8 heap, has 45% malloc fragmentation, and a peak footprint of 746 MB that never gets released, classic leak pattern.
On my 16 GB Mac, a "normal" workload (2 Claude sessions + browser + terminal) pushes me into 9.5 GB swap within hours. My laptop genuinely runs slower with Claude Code than when I'm running local LLMs.
I get that shipping fast matters, but building a CLI with React and a full Node.js runtime is an architectural choice with consequences. Codex proves this can be done in 15 MB. Every Claude Code session costs me 360+ MB, and with MCP servers spawning per session, it multiplies fast.
First: marginal inference cost vs total business profitability. It’s very plausible (and increasingly likely) that OpenAI/Anthropic are profitable on a per-token marginal basis, especially given how cheap equivalent open-weight inference has become. Third-party providers are effectively price-discovering the floor for inference.
Second: model lifecycle economics. Training costs are lumpy, front-loaded, and hard to amortize cleanly. Even if inference margins are positive today, the question is whether those margins are sufficient to pay off the training run before the model is obsoleted by the next release. That’s a very different problem than “are they losing money per request”.
Both sides here can be right at the same time: inference can be profitable, while the overall model program is still underwater. Benchmarks and pricing debates don’t really settle that, because they ignore cadence and depreciation.
IMO the interesting question isn’t “are they subsidizing inference?” but “how long does a frontier model need to stay competitive for the economics to close?”
Earlier quoted context omitted.
We don't vary our model quality with time of day or load (beyond negligible non-determinism). It's the same weights all day long with no quantization or other gimmicks. They can get slower under heavy load, though. (I'm from OpenAI.)
Specifically including routing (i.e. which model you route to based on load/ToD)? PS - I appreciate you coming here and commenting!
Earlier quoted context omitted.
One aspect of this is that apparently most people can't draw a bicycle much better than this: they get the elements of the frame wrong, mess up the geometry, etc.
There's a research paper from the University of Liverpool, published in 2006 where researchers asked people to draw bicycles from memory and how people overestimate their understanding of basic things. It was a very fun and short read. It's called "The science of cycology: Failures to understand how everyday objects work" by Rebecca Lawson. https://link.springer.com/content/pdf/10.3758/bf03195929.pdf
The bicycle frame is a bit wonky but the pelican itself is great: https://gist.github.com/simonw/a6806ce41b4c721e240a4548ecdbe...
One aspect of this is that apparently most people can't draw a bicycle much better than this: they get the elements of the frame wrong, mess up the geometry, etc.
> We build Claude with Claude. Our engineers write code with Claude Code every day well that explains quite a bit
CC has >6000 open issues, despite their bot auto-culling them after 60 days of inactivity. It was ~5800 when I looked just a few days ago so they seem to be accelerating towards some kind of bug singularity.
To me, their claim that they are vibe coding Claude code isn’t the flex they think it is.
I find it harder and harder to trust anthropic for business related use and not just hobby tinkering. Between buggy releases, opaque and often seemingly glitches rate limits and usage limits, and the model quality inconsistency, it’s just not something I’d want to bet a business on.