
Last Week in AI #231 - Claude Cowork, Anthropic $10B, Deep Delta Learning
541 snips
Jan 21, 2026 Anthropic unveils a new cowork tool integrating Claude Code, promising efficiency across various tasks. The company raises a staggering $10B, highlighting the booming AI startup market. Meanwhile, Nvidia struggles to meet the demand for H200 chips amid supply challenges from China. Significant discussions arise surrounding U.S. export controls, with key figures debating the future landscape of AI. Google faces backlash after removing flawed AI health summaries, raising safety concerns about technological integration in personal health.
AI Snips
Chapters
Transcript
Episode notes
Desktop Agents With Sandboxed Safety
- Anthropic's Cowork bundles Claude Code-like automation into a desktop agent that manipulates files and apps.
- It runs models in sandboxed VMs to retain guardrails while enabling powerful local tasks.
Budget Agents As Labor, Not Just Tools
- Treat high-tier agent subscriptions as outsourced labor rather than simple autocomplete.
- Expect $100–$200/month pricing for heavy agent use and plan budgets accordingly.
Personalized Reasoning Inside Gemini
- Google’s Gemini personal intelligence can connect Gmail, Photos, Search, and YouTube for reasoning over personal data.
- The feature is opt-in with guardrails for sensitive topics to limit over-personalization.
