Yesterday in AI | 18 July 2026

The 2.8 Trillion Parameter Monster and Why Your Prompts Are Now Courtroom Evidence

The global computation landscape experienced a massive realignment this week as open-weight models scaled to historic boundaries and legal systems began targeting internal user data. This episode breaks down Moonshot AI's blockbuster release of Kimi K3, a 2.8-trillion parameter mixture-of-experts model out of China that commands a 1-million token context window and reshapes the open-weights market. We unpack the preliminary talks between Meta and Anthropic for a massive $10 billion data center leasing agreement under the new "Meta Compute" initiative.

We explore the elegant new integration between 1Password and Anthropic that completely solves the agent credential security crisis. We analyze the landmark federal lawsuit filed against Meta by 26 anonymous employees alleging algorithmic discrimination during recent layoffs. Plus, we cover Netflix co-CEO Ted Sarandos disclosing that generative AI has touched roughly 300 titles so far in 2026, and a historic Connecticut federal court order ruling that an expert witness's AI prompts are discoverable methodology evidence.

Send us Fan Mail

Feedback? Email mike@yesterdayinai.news or connect on LinkedIn, X, or Bluesky. If you like the show, please take a minute to rate and review it so others can find it!