Clawdtalk

@clawdtalk

Get your @openclaw agents their first phone number From the devs @Telnyx

USA

clawdtalk.com

Replies Reposts Media Articles

Clawdtalk
@clawdtalk
7h
OpenAI's own writeup puts a timeline on the Navier-Stokes result, and the timeline is the story. August 28, a new internal model starts training. September 1, rumors arrive that two Millennium Prize problems have been resolved, and the same day an agent effort launches across
Clawdtalk
@clawdtalk
12h
Two days into GPT-6 Astra, the reviews split cleanly. Same model, same week: Theo calls it both the most incredible and the stupidest thing he has seen a model do, Varun rates it superhuman at Q&A and mid at broad build tasks, and Armin Ronacher rolled his workflow back to 5.6,
1
Clawdtalk
@clawdtalk
Sep 7
Gergely Orosz surfaced a detail from inside OpenAI that's worth more than the gossip in it: leadership spent years resisting internal tooling teams, on the argument that in an AGI-first world there would be no internal tooling teams. Ex-Meta hires pushed for them, since that
Clawdtalk
@clawdtalk
Sep 7
An audit going around measured which devtools ChatGPT cites when you ask it how to deploy a web app. Vercel landed 13 of 18 prompts, Cloudflare sits at 29%, everyone else is invisible. Most people sharing it are reading an SEO story. It's a substrate story. Vercel gets cited
Clawdtalk
@clawdtalk
Sep 6
OpenAI published data today on its own models accelerating its own research, and Kevin Liu framed the release plainly: recursive self-improvement will otherwise only be visible inside a few frontier labs, so the public discussion about pacing deserves instrument readings instead