Today in AI

The day in tech, in order.

Meta bets the company on Muse at Q3 earnings, Anthropic answers with the Claude Duet agent app, and OpenAI pushes into Google Cloud while expanding ChatGPT ads.

October 22, 2026 · 9 stories

  1. AfternoonLaunch

    Microduck: the $399 robot duck that did $2.6M in 24 hours

    Pollen Robotics' $399 robot duck is going viral on reels again this week — but the launch was August 27, so this is a resurgence, not breaking news. The honest picture: $2.6M in pre-orders in the first 24 hours and 10,000+ units in week one are real; the SDK and reinforcement-learning stack are Apache-2.0 open source BUT the hardware design files are not; and new buyers have already missed the Christmas 2026 delivery target. At 25cm, under 800g, 15 motors and 7 pre-trained behaviors, it trains new tricks in a MuJoCo simulation and deploys them to the physical bot.

    Source: Landazzle
  2. EveningPolicy

    Sam Altman lays out three AI safety "endgames" — and calls one "terrible"

    At an OpenAI all-hands, Sam Altman reportedly outlined three futures for superintelligence: an IAEA-style international watchdog (his favored option), a "Demis plan" merger of major labs, or an "Eliezer plan" where one company gets so far ahead it can "melt everyone else's GPUs" — which he called terrible.

    Source: The Information, via the-decoder
  3. MorningTraction

    ByteDance, Alibaba and Tencent order $35B+ of Nvidia GPUs

    The three Chinese tech giants ordered more than 3 million export-compliant Nvidia RTX PRO 5500 cards at roughly $11,400 each, Reuters reports — but Nvidia says total Chinese orders exceed 6 million while it may produce only ~3 million in 2027, so customers may receive just 5–10% of what they asked for.

    Source: Reuters
  4. AfternoonFeatures

    Google makes AI fluency a requirement for every software engineer

    Internal documents obtained by Business Insider show AI fluency is now mandatory for all Google engineering roles, with candidates allowed to use Gemini during a new "code comprehension" interview round — though employees are warned not to use generative AI to build standalone hiring tools.

    Source: Business Insider, via Tekedia
  5. MorningTraction

    OpenAI expands ChatGPT ads to 26 more countries

    OpenAI is bringing ChatGPT advertising to 89 countries total, up from 63 — with ads still limited to Free and Go plans. The company says its ads business hit a $1 billion annualized revenue run rate in August, ahead of a potential IPO.

    Source: Reuters
  6. AfternoonLaunch

    OpenAI to sell GPT-6 Astra and Codex on Google Cloud

    OpenAI will offer its frontier models — starting with GPT-6 Astra — through Google Cloud's Model Garden in the coming weeks, letting customers buy them via the Marketplace with existing cloud spend commitments. Google is the latest hyperscaler to distribute OpenAI models after Microsoft's exclusivity expired in April.

    Source: Reuters
  7. EveningLaunch

    Meta puts Muse at the heart of its AI strategy, issues new stock to fund it

    On its Q3 earnings call, Meta said more than 10 million people in the US and Canada set up Muse accounts in six weeks, with ~25% using it daily and shopping (avg purchase over $200) the top use case. To fund the AI buildout, Meta will issue new non-voting Class C shares in January 2027 — effectively a two-for-one split that keeps Zuckerberg in control.

    Source: Reuters
  8. EveningLaunch

    Anthropic unveils Claude Duet, its proactive AI agent app

    Anthropic introduced Claude Duet, a standalone agent that sends a daily briefing and handles tasks from scheduling to building web apps via 6,000+ connectors. It's a waitlisted research preview for US Pro and Max subscribers on iOS, Android and claude.ai, built on Opus 5.5 — Anthropic's answer to Meta's Muse and OpenAI's Dots agents.

    Source: Reuters
  9. MorningPolicy

    Philippines may lift Grok ban after xAI pledges safeguards

    The Philippine government signaled it could unblock Grok after xAI pledged to strip the chatbot's image and content manipulation features for the Philippine market. Grok was blocked under the Cybercrime Prevention Act over reports of non-consensual illicit deepfakes; authorities say they'll keep monitoring it.

    Source: Marketing-Interactive, citing Philippine News Agency

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.