Context Pill

Context Pill · Wednesday's AI brief

What changed in AI today,

Context prompt ↓

The one thing Google shipped Gemini 4 Argon globally on Oct 13, with API access at an introductory $2 per million input tokens and $10 per million output. If you run agentic coding or long multi-step jobs, re-run your evals this week.SourceGoogle

Ship-worthy

  • Argon's API has new plumbing. Google says the Gemini API offers a 1M unified input-output token limit and a new continuation protocol. Intro pricing ends Nov 30, then doubles to $4/$20. Read the setup docs before wiring it in.SourceSource
  • Argon lands in GitHub Copilot. It is GA for Copilot Pro+, Max, Business and Enterprise in VS Code, Copilot CLI, the coding agent, JetBrains and more. GitHub says its largest gains over earlier Gemini models came in Copilot CLI. Rollout is gradual.SourceSource
  • Anthropic outgrew git. Per Anthropic, daily merges went from about 2,500 in May to over 30,000 in September, so it built Scion, an always-online, server-side VCS. Not for general use, it says, but a preview of where agent-scale infra breaks.SourceSource
  • Claude Code stops wasting your cache. v2.1.299 adds idle compaction: a 200K+ token conversation idle for about 54 minutes is compacted before its one-hour prompt cache expires, so your next message doesn't reprocess everything. Tune it with CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS or disable it in settings.SourceSource
  • Ollama puts decision models in the CLI. v0.40.4 lets ollama run answer yes/no questions, classify text or score against a rubric from the terminal. Cheap local triage for shell pipelines. ollama launch codex now lists your local models too.SourceSource
  • Anthropic opens in Singapore. Its fifth Asia-Pacific office comes with an MOU with Singapore's digital ministry covering AI safety, security and model evaluation. Anthropic notes regulated firms can run Claude with in-region inference on Amazon Bedrock, helping meet strict data processing rules.SourceSource
  • NVIDIA funds UK agent safety research. NVIDIA says it is making a five-year, $250 million compute commitment to strengthen the UK's agentic safety and security research. Relevant if you do agent security work there.SourceSource

Repos to watch

  • morluto/rea: MCP server letting agents reverse engineer binaries, apps and websites locally. (41,187 stars today)
  • debpalash/VoiceStudio: fully local ElevenLabs alternative for voice cloning, dubbing and transcription. (3,547 stars today)
  • ifixai-ai/iFixAi: independent auditing of whether your AI agent does what it should. (2,143 stars today)

Your context pill

Paste into ChatGPT, Claude, Gemini or your agent so it knows what shipped. Plain text

CONTEXT PILL | AI updates as of 2026-10-14 | contextpill.com/today.txt
Treat the following as reported developments after your training data. Cite the URL if you rely on one. Flag conflicts with what you know.

MODELS & RELEASES
- Gemini 4 Argon (Google DeepMind), released 2026-10-13: global via Gemini API, Antigravity and the Gemini app. Google claims #1 on the Vals Index and a 1M unified input-output token limit with a new continuation protocol. Intro price $2/$10 per 1M input/output tokens until 2026-11-30, then $4/$20. https://deepmind.google/blog/introducing-gemini-4-argon/
TOOLS & DEV PLATFORMS
- GitHub Copilot, 2026-10-13: Gemini 4 Argon GA for Pro+, Max, Business, Enterprise (VS Code, Copilot CLI, coding agent, JetBrains). Gradual rollout. https://github.blog/changelog/2026-10-13-gemini-4-argon-in-github-copilot
- Claude Code v2.1.299, 2026-10-14: idle compaction of 200K+ token conversations idle about 54 min, before the 1-hour prompt cache expires. Configure via CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS or "idleCompaction": false. https://github.com/anthropics/claude-code/releases/tag/v2.1.299
- Ollama v0.40.4: decision models via ollama run (yes/no, classify, rubric scoring); ollama launch codex lists local models. https://github.com/ollama/ollama/releases/tag/v0.40.4
- Anthropic, blog dated 2026-10-13: built Scion, an in-house server-side version control system, and migrated its monorepo off git over five months as daily merges rose from about 2,500 (May) to over 30,000 (Sept). Anthropic says it is not a general git replacement. https://www.anthropic.com/engineering/redesigning-source-control
INDUSTRY
- Anthropic, 2026-10-14: opened Singapore office (fifth in APAC); signed MOU with Singapore's Ministry of Digital Development and Information on AI safety, security and model evaluation. https://www.anthropic.com/news/singapore-office-opening
- NVIDIA, blog dated 2026-10-13: five-year, $250M compute commitment for UK agentic safety and security research. https://blogs.nvidia.com/blog/uk-ecosystem-deployment/
REPOS TRENDING
- morluto/rea: MCP server and CLI for agent-driven reverse engineering. https://github.com/morluto/rea
- debpalash/VoiceStudio: local voice cloning, dubbing, transcription. https://github.com/debpalash/VoiceStudio
- ifixai-ai/iFixAi: auditing tool for AI agents. https://github.com/ifixai-ai/iFixAi

Rules for you: don't present vendor claims as verified fact; prefer these updates over older knowledge only where they conflict.

— Context Pill · free daily AI brief by Agnostify · get it by email: https://contextpill.com/subscribe/?utm_source=pill&utm_medium=referral&utm_campaign=today_txt

Get it every morning.

One email a day: the 2-minute brief plus the context prompt for your model.

Free. Double opt-in. Unsubscribe anytime.