This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
Enough has already been said about errors of conservatism in this post. I feel bad that I'm not in the shape to offer counter-predictions. One detail.
5.3 Codex is exceeded in its direct niche by a whole lot of Chinese models, credibly at least by Qwen 3.7, and definitely by Composer 2.5 (Cursor/xAI) which is a finetune of Kimi K2.5, itself a continued pretrain of Kimi K2, which is a year old base model using a two year old DeepSeek architecture.
Things are moving very fast, and the gap between China and the US, and frontier and open source, is measured in months, not years – at least on main axes of comparison; but these months contain a lot of distance (5.5 is way stronger than 5.3-Codex). I think you don't understand what's happening here. With RLVR (and its generalized form – terminal feedback, all program execution traces), we have an infinite source of ground truth. We can just have models try arbitrarily complex things and reinforce what has worked. We can have them strategize of how to do it and it won't be useless. We have the compute for enormous waste. We have absurdly capable and efficient models (eg said DeepSeek charges $0.003635 per 1 million cache hits, and they probably price slightly above cost; it is hard to imagine that the Western frontier doesn't have anything close, after all these techniques are openly published). So RL will keep improving models unexpectedly fast, by the end of 2026 we'll be discussing the perils of open sourced Mythos, not Codex. Sorry, this is reality, we're not close to any S-curve plateau. We won't have the time for this neat slowdown to do "theory".
I've addressed these claims elsewhere in this thread. My question for you is, do you even use these models? My experience with Chinese models vs. OpenAI models does not align with the benchmarks. I don't trust the benchmarks.
We'll see. I can't see how your predictions are superior to mine objectively. Maybe they come to fruition but I'm not seeing evidence for it at the moment.
Yeah, I don't mean only the benchmarks either. Codex 5.3 is just behind the curve. If you think catching up to it can take up to two years, then I guess your idea of what makes Codex capable has to do with very superficial short-horizon polish.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link