Evening brief

Tuesday, July 7

3 people · 0 companies · 6 topics

TL;DR Fable 5 exits subscriptions tonight and moves to pay-as-you-go credits from July 8. GPT-5.6’s general access window opens this week per Polymarket ($29M+ volume, 99% odds by July 31), with the White House voluntary AI standards framework as the likely trigger.

Models & research

  • model-releases Fable 5 billing cliff: today is the last day Fable 5 draws from subscription limits on Pro/Max/Team/Enterprise plans; from July 8 it bills at API rates ($10/1M input, $50/1M output). Anthropic says the move is temporary pending capacity. Bleeping Computer · Anthropic post

  • model-releases GPT-5.6 Sol/Terra/Luna still government-gated but the general access window opens this week (July 7-14); OpenAI said it’s working toward broad release after the White House completes voluntary AI standards framework. Polymarket: 99% yes by July 31, $29.3M traded. OpenAI preview · VentureBeat · (Polymarket: 99% yes, $29M volume)

  • model-releases claude-code Claude Sonnet 5 is now the Claude Code default — 1M-token native context, promo pricing $2/$10 per Mtok through Aug 31, bumped to $3/$15 after. Released June 30. Anthropic · Releasebot

  • model-releases Anthropic in early talks with Samsung to manufacture a custom AI chip on Samsung’s 2nm process with advanced packaging; no design yet, but an OpenAI chip engineer reportedly just joined. TechCrunch · Bloomberg

  • ai-research SimFoundry (NVIDIA GEAR + drjimfan’s lab, plus Fei-Fei Li’s team): Real2Sim system that turns a single real-world video into an interactive robot sim environment; policies trained in it transfer zero-shot with 17-40% success rate gains. arXiv 2606.28276 · NVIDIA research

People

  • karpathy shared highlights from his Sequoia Ascent 2026 fireside chat: “vibe coding raised the floor, agentic engineering raises the ceiling.” He called December 2025 the tipping point when agentic coding tools started consistently producing correct code, and flagged three new LLM use-case categories beyond speed-ups (e.g. LLM-generated knowledge bases). X post · blog

Tools & dev

  • claude-code Claude Code reliability push: recent update switches the default permission mode to “Manual,” fixes ~2.7s Remote session startup lag, fixes CJK text pasting as mojibake, and improves screen reader accessibility. Releasebot

  • dev-tools Cursor Developer Habits Report: 30% of PRs now auto-submitted by AI agents, cloud agent usage up from ~17% to 75% since Jan, and 60%+ of code is AI-generated. University of Chicago study found companies merge 39% more PRs after adopting the agent mode. Cursor Insights

OSS

  • oss Kimi K2.7 Code in GitHub Copilot (GA July 1): first open-weight model in Copilot’s model picker; 1T total params, 32B active per token, hosted on Azure. Available to Pro/Pro+/Max plans; Business/Enterprise rollout coming. GitHub Changelog

  • oss system_prompts_leaks by asgeirtj trending on GitHub today: 51k stars, 1,378 added today — a regularly updated collection of extracted system prompts from Claude, GPT, Gemini, Grok, and others. GitHub

Other

  • UN Global Dialogue on AI Governance (Geneva, July 6-7) and ITU AI for Good Summit (July 7-10) are running concurrently; 193 member states at the UN dialogue, 170+ countries at ITU. Focus areas include safety standards, open-source governance, and interoperability. UN · ITU