In my experience this "gaming" behavior is easily caught by just asking another agent (could just be another session of Claude Code) to review the code changes.
For idle sessions I would MUCH rather pay the cost in tokens than reduced quality. Frankly, it's shocking to me that you would make that trade-off for users without their knowledge or consent.
That's true, but the "AI bubble bursts" scenario is usually tied to Western investors getting essentially margin-called. If that happens, the CCP won't suddenly stop their investment; Chinese models will most likely continue developing.
> I have no clue if this claim holds, but alas, just pretending they did not address the obvious criticism, while they did, is at the very least pretty lazy.
But they didn't address the criticism. "cutting ~75% of tokens while keeping full technical accuracy" is an empirical claim for which no evidence was provided.
> Aren't they currently propped up by investor money?
Are Chinese model shops propped up by investor money? Is Google?
Open weights models are only 6 months behind SOTA. If new model development suddenly stopped, and today's SOTA models suddenly disappeared, we would still have access to capable agents.
These OSS model makers need to stop benchmarking against old models. Showing how it performs against Opus 4.5, GLM-5 when we have Opus 4.6 and GLM-5.1 just tells me that it's not comparable to SOTA.
I think there's a lot of methodological expertise that goes into collecting good eval data. For example, in many cases you need human labelers with the right expertise, well designed tasks, well defined constructs, and you need to hit interrater agreement targets and troubleshoot when you don't. Good label data is a prerequisite to the stuff that can probably be automated by the AI agent (improving the system to optimize a metric measured against ground truth labels). Data scientists and research scientists are more likely to have this skillset. And it takes time to pick up and learn the nuances.
Well you're free to disagree but my experience has been counter to your position. I write both code and research / technical documentation. The quality of what the LLM produces is limited by the quality of ideas I give it initially (mind you, this is just a starting point), and the quality of my review of its output.
I haven't seen the scrolling glitch in months, where previously it was happening multiple times a day. Also haven't seen anyone complain about it in quite some time. Pretty sure they have resolved that.
> I think the only reason it’s seen as good anywhere is there are a lot of tasteless and talentless people who can pretend they created whatever was curled out. This goes for code as well.
This is an oversimplification.
If you have taste and talent, then the LLM output you get is going to reflect that.
So on the one hand, yes: tasteless and talentless people won't know good output from bad output. On the other hand, people with taste and talent can actually get good output.
This couldn't be more wrong. The simplest refutation is just to point out that there are temperature and top-k settings, which by design, generate tokens (and by extension, ideas) that are less probable given the inputs.
Vegan for 15 years. I cook 95% of my own meals, including black bean burgers, tofu, etc... Sometimes I want something that tastes like meat and I reach for a Beyond or Impossible burger. I don't need it. But I can't recreate its texture and flavor profile on my own. It's not "better" than other things I can cook. It's just different.
But literally any decent agent can recommend existing services and help you set them up. And even help you help them set the services up for you. I do this with Claude all the time.