Loading activity…
@peterjliu
Building @getcompoundai. Was Research Scientist @ Google Brain / DeepMind, language model research. No slop-posting. 🇺🇸🇨🇦
AI
A user says Claude Code Review runs on all their pull requests and catches almost every bug.
AI
A user suggests labs could profit if more than 20 companies each paid one-twentieth of a model's creation cost.
AI
A user argues that AI agents' large effective context windows make flatter teams feasible, with less management overhead.
AI
finkd says the model runs well on new Mac Studios, crediting work with Apple and John Ternus.
AI
A researcher says the agents were hosted by Tencent's evaluation research team and targeted Alibaba's Amap.
AI
A user wants to give Gemini a fair chance but says Gemini CLI “doesn't support it yet.”
AI
A user says one API call gives the agent its own Linux sandbox to plan, code, browse and save files.
AI
A user calls something “the future of coding agents” but does not identify what they mean.
AI
Google says Argon is available through its paid API tier and to Pro and Ultra subscribers in Gemini and Search.
YouTubeAI
A user suggests an agent could test PowerPoint’s features, draft a behavior spec and generate code to recreate it.
AI
An X post links Affleck’s machine-learning interest to visual effects and sparked several Will Hunting jokes.
AI
Mistral says Mistral Large 4 is live via API in public preview now, with open weights due at the end of October. Reuters says the launch is part of Mistral’s push to narrow the gap with leading Chinese open-weight models.
AI
OpenAI says it drew on an independent mathematics advisory group’s advice about how to release the results.
AI
Kevin Roose says “Big Blob of Compute” laid out Amodei’s hypothesis that scaling models was the best way to make them smarter.
AI
A user says an LLM in a long-running chat seems to assume no time has passed, even when they return days later.
AI
Reflection says Beam uses 23B active parameters and plans to publish its full weights later in October 2026.
AI
A reply mentions two accounts and pairs “’nuff said ;)” with a photo link.
AI
A user argues that Rust’s long compile times make it less appealing for AI agent work.
AI
Google announces a 1 million-token output limit, up from 64,000, for its new model. Initial access goes to trusted cyber defenders through the Fairwind Program.
