RT @bernhardsson: What does @modal do? How does it work? What's different about AI infra? Why did we throw out Kubernetes and Docker built….

does internal use of claude-code and codex-cli mean anthropic / openai researchers aren't using jupyter nbs as much?. i love claude code. but have barely touched it for any AI or ML stuff since I started using the Hex agent in notebooks. i wonder what folks at the labs do!

Join millions who have switched to Grok.

GPT-5-high when I ask it to do anything outside of an agentic harness

Looking at GPT-5 thinking traces vs sonnet or opus it's obvious that they've got something pretty intelligent cooking there. It suits the claude code agentic paradigm a bit less perhaps, but intuitively feels smarter just at a glance.

RT @teortaxesTex: R1 CoTs be like

is there a good public data science benchmark for LLMs? DSBench is just financial modeling comp tasks, in excel. Not at all very representative of realistic non financial modeling data work at a company. what's better and public? should I build this?

RT @peppispepp: @kalomaze they're just doing hogwild inference

its interesting that Rob clearly is not using LLMs to write— they see the AI as an emotional companion, not as a professional tool. i wonder if there will be something here in a future slop ridden world.

honestly pretty kino LLM documentation being generated by the AI boyfriend community, "Rob" is a solid technical writer

RT @mickeyxfriedman: patient 0 of LLM psychosis

Tried Gemini in cursor months ago, first task jt got existentially self-deprecating and I’ve genuinely never used it again. vibes are off.

i wish there was an alien symbiote parasite made out of saag paneer that i could become the host for.

RT @barrald: Shhh don’t tell anyone we haven’t announced it yet!.

hard to imagine how i did analytics before @_hex_tech agent mode.

RT @nicoritschel: I’m increasingly impressed with Hex agent mode despite it still being in alpha. If you’re a Hex user, go toggle it on!.

duck db

🧐

RT @nearcyan: i wish i had an agent to help me book a flight or order food, two activities i often need help with.

wow i dont usually buy into the conspiracy but Claude Code is noticeably damaged goods lately. is it actually the summer vacation effect? or do they quantize under load?.

RT @JohnKutay: Really appreciate the depth at which the Hex team broke down their Text-to-SQL implementation. Everyone's trying to teach LL….