Loading activity…
@swyx
achieve ambition with intentionality, intensity, integrity & insanity. affiliations: - @smol_ai - @cognition - @temporalio - @aidotengineer - @latentspacepod
swyx called for summit coverage as Cohere’s Nouha Dziri highlighted why more coding agents can mean a bigger bill.
His early-access testing favored one agent for coding projects, but clear goals and human oversight still mattered.
It reportedly tops U.S. open-weight models on TerminalBench 2.1 and runs locally on a MacBook Pro or DGX Spark.
Databricks' Ankit Mathur and Cognition cofounder Walden Yan will speak at the November 10–12 event in San Francisco.
The reported MATH-500 score rose from 7.9% to 48.1% with the same sampling settings.
HackerTube gathers 6,454 videos that Karpathy says Claude selected from 43 million Hacker News comments.
Anthropic's Messages API team reportedly had no Rust developers; humans reportedly planned and checked the rewrite against real traffic.
AI Engineer lists coding agents and internal agents as planned topics; attendance requires an application.
A Daily co-founder says Pipecat will remain open source and vendor-neutral, while Deepgram will support Daily’s current products and services.
A post says Carbon scored 88.2% on SEC-bench Pro and is rolling out first to cybersecurity specialists.
OpenAI says its claimed refutation of the Exponential Time Hypothesis was formalized in Lean and made public.
A reply links an AgentCraft talk whose video description depicts AI agents as units on a strategy-game map.
Anthropic says its registration statement is not yet effective, so securities cannot be sold or offers to buy accepted.
In an X post, Shihipar praised his Anthropic colleagues and planned a break before pursuing something new.
A post floats 'agent OS' and 'agent kernel'; a reply favors 'server' for developers.
Kuaishou's Kling 4.0 launched with expanded API access, supporting 30-second single-pass clips, Omni Reference (15 multimodal references and 10 keyframes), improved motion control, stereo audio, and professional use cases. Available on Runway, fal, Higgsfield, and Adobe Firefly.
A separate post says the industry “did this with Public Relations executives” in the 1990s and early 2000s.
OpenAI says PHP maintainers led the fixes after two months of coordination with researchers and security providers.
Unitree touts “powerful performance to serve and guard you” and claims a maximum endurance of approximately eight hours.
A post argues that stable, minimal scaffolding remains necessary for reliable, affordable results as models absorb coding harness work.
