OpenAI's streak of aura losses continue. Lock in! 😭

~90% of frontier lab compute now goes to post-training and inference.

no there is no fucking way tales of using AI "please make a dashboard where I can see my balance" "dashboard ready! access it at: http : // . . ." "wow that's beautiful! but where's the balance?" "just type it, and it will appear!" - Opus 5.5

Why I’m alarmed by Anthropic, in the @nypost: nypost.com/2026/10/06/tech/a…

"do you literally type like that?" yes, this is how I work now I sit down, open ????Chat, type my stream of thoughts out for 30 minutes, press enter, and wait it get done below I teach it about Bend's round, while it builds its own app this is my life now I miss VIM though

Four updates or a reset. Or both. How was day 2.

We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math

Le Chaton Fat is real lol. > “It is the best open weights model from US or Europe on aggregated benchmarks.” Europe OSS being ahead of US OSS LLMs was not on my bingo card. they claim it beats GLM-5.3.

It's pretty funny that it seems the one thing Mistral Large 4 is actually leading on is a benchmark about regulation Another EU banger

You simply get another girl who plays Valorant. They all look like this.

What the fuck even is valorant?

The single most useful take I heard at The Curve was "we can still heist the Mona Lisa after the singularity". There's a subtle/important point here, where some things (e.g, n of 1 goods where originally is valuable) will retain/grow in value even as everything changes.

any of the following can and will give women the ick: * emojis, “haha”, “lol” * any sort of sexual innuendo especially if you initiated it * replying too quickly * any sort of “good guy” behavior * any words that aren’t already backed up by actions i can keep going

I still cannot believe something that dumb is transforming my life to this extent, I spent the day happy and I'm still happy it feels like I'm running out of cool ideas because I just keep throwing tasks at it and it keeps going

I'm not even talking about work. I'm suddenly solving all these small problems on my life that I couldn't find the time to do. like buying a new controller for my TV, renewing old domains, migrating Digital Ocean servers. I just describe a problem to it, and wait it get solved. I've been using the same chat for days, and it just keeps going. I built an entire app in 2 days, deployed its servers, implemented a CLI harness, a website, mobile, and all devices sync to the same chat via websockets. there are payments, animations, the site is beautiful. I can use it to control my macbook from the shower now. and all of this was built without a single /compact. and then I've booked flights, researched EU customer laws, rent GPUs. all from the same chat. and the AI doesn't get confused. there is no context rot, because the context never grows large. it just... keeps doing. to be clear, these capabilities are not new. the intelligence was always there. models since GPT 5 or so were already smart enough to solve basically any mundane problem that you could have. but *organizing* all these concurrent agents and blobs of knowledge took some human time that just didn't allow me to scale and use the models to their full potential. now, I feel like there is no limit to how much knowledge the agent can absorb, and how many agents I can manage at a point, specially during Bend's development, I was managing 10+ concurrent agent, each with a different prompt, managing a different part of the project, and I couldn't manage all the chaos and complexity. I couldn't just put all of bend, let alone my life, in a single context, because that'd be a million tokens long. so, each of these agents had to have different contexts, prompts, memories, each covering a different sub-domain of the whole. and they sucked at managing it, so I had to. they would delete important info, save wrong info, over-compress or under-compress. memories would get stale, or grow too large. I needed to be there holding their hands, managing the chaos, moving blobs of knowledge around, re-explaining insights in different contexts. that prevented me from scaling and it was miserable. "now write all you learned to memory.txt, prune out the stale bits about the computed lambda campaign, and preserve the bits about the operator elaboration and style rules. ensure the file is still under 64k ttok" and then it would still fucking forget now I can actually literally dump all of Bend in it, let it build the memory tree, and the information will just be there, in a single place, easily accessible. I never need to re-explain something again, I just tell it to seek the info I already written once. because it is there, there is a quick navigable path to it. or to anything else in my life this is just so useful from a human perspective. I feel like I'm running out of mundane problems to solve now. I was literally walking around the house, looking at stuff and thinking "what else can I fix"? because my time is free, I have a problem-solving hammer and I wanna use it. that's all it takes to get a problem solved now: remembering it exists, and throwing it at the problem-solving information sponge sorry I will stop, I should log out, I'm just feeling a sense of freedom and simplicity that is beyond what I can put in words

man i can’t wait for this chat. oh man

My last day on the job

I haven’t touched Claude Code or Codex CLI in a while. The terminal era is over imo. It's the wrong interface for coding agents. Tabs are ephemeral, but context is persistent, and managing 30 tabs is pure cognitive overhead. I don’t really need an IDE like Cursor either. I rarely navigate the whole codebase anymore. The new primitive is the agent, not the file. (Codex desktop app is the best agentic UI for now. But we’re still early.)