Today in AI

The day in tech, in order.

Meta's first frontier models lead a packed stretch: an X subscription bundle, Anthropic's anti-distillation push, and Nadella's call for AI 'emergency brakes'.

October 13, 2026 · 5 stories

  1. Evening ISTLaunch

    Meta Launches Muse Wonder and Muse Wonder Full, Its First Frontier AI Models

    Meta's Superintelligence Labs released its first frontier models with 'Active Memory' (context across conversations) and 'Active Focus' (persisting through hard tasks). Wonder hits the Meta Model API at $4/1M input tokens; Wonder Full is Muse Code-only. Independent scores place Wonder Full close behind Opus 5.5. Meta shares rose ~4%.

    Source: Reuters
  2. Evening ISTTraction

    SpaceXSI Folds Grok, Grok Bot, Cursor and X Premium Into One "X Subscription"

    Five tiers from free to $200/month share one Grok usage pool across the Grok apps, X, and Cursor's coding agents, replacing the fragmented SuperGrok, X Premium and Cursor plans. Rolling out globally over the coming days.

    Source: x.ai
  3. Morning ISTSecurity

    Anthropic Says New Safeguards Cut Claude Distillation Attacks 90–99%, Publicly Names China's Z.ai

    Reasoning-token extraction from Claude is now 10-100x lower, Anthropic says, after banning tens of thousands of accounts. It publicly accused Chinese startup Z.ai (Zhipu) of distilling Claude's capabilities for its GLM-5.3 model. Figures are Anthropic's, not independently verified.

    Source: Reuters
  4. Oct 10 · EveningPolicy

    Satya Nadella: Treat AI Models as Compromised — Every Agent Needs an "Emergency Brake"

    Microsoft's CEO laid out seven trust principles in an X post: assume a model is compromised from the start, keep tamper-proof records of agent actions, and make sure an authorized human can pause any model mid-task.

    Source: TechCrunch
  5. Morning ISTTraction

    Karpathy: only about 5,000 people have seen what frontier AI agents can truly do

    Anthropic's Andrej Karpathy re-shared his four-tier funnel: ~6B people barely touched LLMs, 20M professionals use frontier models daily, and ~5,000 inside labs watch agent swarms work for weeks. His warning: human review is starting to lag machine output.

    Source: Business Today

Timelines are composed each morning from the previous day's verified news — every story links to the source it was checked against. See the full feed on the news page.