Daily Digest

OpenAI's new flagship family, "frontier intelligence that scales with your ambition," live in ChatGPT/Codex/API; Microsoft 365 Copilot already defaults to it

New agent in ChatGPT that takes action across apps and files, powered by Codex + GPT-5.6

Yesterday's voice model is now live for all Go/Plus/Pro users, free users in progress

Today's top HF paper (96 upvotes)

Emil Kowalski distilled Apple's WWDC videos into 17 design + motion principles as a skill; 9.3k likes

A browser video editor built to be driven by AI agents, zero deps

Brown professor's in-person finals exposed take-home AI use; 32k likes, big debate on assessment in the AI era

Simon Willison: "so confused by ChatGPT v. Codex v. ChatGPT Work v. Claude v. Claude Code v. Claude Cowork"; Ethan Mollick echoes on Cowork vs Code

Ethan Mollick calls out xAI: a serious AI company should ship a model card

Ben Bernanke joins the Long-Term Benefit Trust — Former Fed chair added to Anthropic's governance body | @AnthropicAI

Reflect with Claude — New feature to review how you use Claude (Jul 10) | link

GPT-5.5 Bio Bug Bounty — Bug-bounty program targeting biosecurity risks (Jul 9) | link

A universal benchmark for proactive agents on real-world tasks (HF 20)

Semantic runtime auditing for persistent AI agents, watching the token stream to catch anomalies

Persuasion-style attacks reduce chain-of-thought monitoring's ability to catch misbehaving agents

1310 pts, 920 comments, #1 today

When agents generate most code, how does git adapt?

Agent-drivable browser video editor

Analyzing Microsoft's $2.5 Billion AI investment (Jul 10)

Real-time interactive video generation (96↑)

Universal benchmark for proactive agents (20↑)