Rene van Pelt

Kimi K3 is just as 'open weight' as the Grok models. All talk an announcement.

I've had too many GPT tokens the past 2 weeks..

@karpathy Screaming at screens is an implementation strategy and pattern now. Does that mean we can stop looping?

One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.

@SherryYanJiang When a usage limit reset is about to expire, i use extra high

does someone have a use case where they use gpt 5.6 sol extra high or ultra, and find that it's worth it??

@PaulGugAI Good point

@renevanpelt I feel like we are not far off Linux distros being essentially a harness first and PC second.. maybe an appliance like experience that you put on a mini PC

I feel like codex, claude code and the other harnesses should less be a harness and more be an operating system.

So how do I use fable? Been weeks, still getting the guardrails stuff too much

@rezoundous Mine have been conspiring to turn against me.

apparently even using GPT-5.6 + Fable for a long time in 1 single session makes it an echo chamber. both getting tunnel vision lol

@svpino Most people didn’t need a car for 99% of their trips a century ago.

The tokenmaxxing culture is absolutely out of hand. You don't need an agent for everything. In fact, I'm 99% sure you don't need an agent for 99% of the problems you need to solve. Good engineering principles still apply. Simplicity still wins. A regular, boring ol' script that works every time is 1000 times better than a fancy agent that works sometimes. We need common sense back.

@mweinbach Graphs are yesterday’s news. It’s stacks now

what the fuck is a graph man just type stuff and hit enter it's not that deep

@rezoundous I think their moment is coming still

How is Google able to fumble this hard despite being so early

@thdxr Selling where?

we banned someone operating 8,000 fraudulent OpenCode Go accounts they were reselling $480,000 of tokens every month this will make things more sustainable for legitimate users enjoy your week

@jun_song Source?

Another fear mongering. ClosedAI is pulling the exact same trick Anthropic did.

@PaulGugAI Apologies, not tweeting on K, posting on X

@PaulGugAI The word lab has absolutely nothing to do with being profit-driven or not. The guy is tweeting on K again

Never considered this perspective. Is this a consensus view on X? I tend to see the term ‘labs’ thrown around more generally..

@elonmusk @huaijiangzhu What do you think a lab means? 😂😂 Same for factories? Farms?

@huaijiangzhu It is pretentious and poisonous for for-profit companies to call themselves “labs” when they are actually corporations and to create a lords & peasants culture between (looksmaxxing) “researchers” & engineers when they are both, on their better days, actually engineers

my spicy theory is that chinese ai labs keep winning because of culture, not talent or resources. chinese labs still have a deeply hands-on engineering culture. deepseek and kimi are flat organizations where science and engineering are fused: the same people move between algorithms, data, and infra, doing whatever it takes to make the model work. but sf tech bros have decided that “researcher” is the high-status title while infra is merely support work. every new sf ai startup calls itself a “lab,” every ambitious engineer quietly upgrades their title to “research engineer,” and the infra work is left to whoever failed to escape it. but at frontier scale, infra determines experiment velocity, and experiment velocity determines research output. at frontier scale, infra **is** research.

@rezoundous Just like cleanliness is with life partners

GPT-5.6's perfectionism is a double edged sword

@PaulGugAI Yeah I had to update Claude Code for it to work. My bad.

@renevanpelt It is (Claude max) but only at 50% of the weekly limit.

So yes or no Fable in the 20x plan? I'm lost

@TimJayas Some orgs in China are making backups for us. Don't worry.

if Anthropic and OpenAI gets disappeared Which will be your backup ai model?