Good evening, AI enthusiasts.

A week ago, investors were debating whether AI stocks were overvalued. Today, they're watching $1 trillion evaporate as AI proves it's not enhancing software - it's replacing it.

In today’s AI debrief:

  • Anthropic triggers $285B software selloff with Claude Cowork

  • Big Tech commits $650B to AI infrastructure

  • Google's Gemini 3 forces OpenAI "code red."

  • Federal vs. state AI regulation battle begins

DISRUPTION

The Debrief: Anthropic released Claude Cowork with plugins for legal, finance, and sales workflows - result: $285B wiped off software stocks in three days.

The details: Claude Opus 4.6 followed with a 1-million-token context window and multi-agent coordination. Financial data providers got hit hardest - FactSet dropped 10%, S&P Global and Moody's saw sharp declines. Markets are betting AI won't just assist software, it'll replace it.

Why it matters: The billable-hour model is under existential threat. The moat legacy software companies thought they had is evaporating. For founders: the question isn't "How do we add AI features?" It's "How do we avoid being replaced by AI entirely?

INFRASTRUCTURE
The Debrief: Alphabet, Amazon, Meta, and Microsoft forecast $650B in capital expenditures for 2026. Alphabet alone: $175-185B, double its 2025 spend. Markets are skeptical despite strong Q4 results (revenue up 18% to $113.8B).

The details: Each company's 2026 budget approaches their combined spending for the past three years. Google Cloud revenue spiked 48% to $17.7B. Gemini app: 750M monthly active users. But investors question when returns will match this massive spending.

Why it matters: This creates a massive barrier to entry. If you're not Google, Microsoft, Meta, or Amazon, you're renting their infrastructure. The opportunity: vertical AI applications solving specific problems, not infrastructure. The picks-and-shovels play is over unless you have $100B.

COMPETITION
The Debrief: Google launched Gemini 3, outperforming rivals on benchmarks and reportedly triggering OpenAI's internal "code red." Combined with a landmark Apple deal, Google's back in the AI race.

The details: Gemini 3 topped key benchmarks. The app hit 750M monthly users. RBC analysts say the momentum justifies the $650B infrastructure spend despite Wall Street's concerns.

Why it matters: The "OpenAI has won" narrative is dead. The model race is wide open. For builders: don't lock into one provider. The landscape shifts monthly. Build for interoperability.

Also worth knowing this week

  • World models rising: Yann LeCun left Meta to start a world model lab seeking $5B valuation. World Labs launched Marble, a commercial 3D-aware AI system for robotics and simulation.

  • MCP becomes standard: Anthropic's Model Context Protocol - the "USB-C for AI" - embraced by OpenAI and Microsoft, donated to Linux Foundation's Agentic AI Foundation.

  • AI recruitment bias: Belgian study found AI tools use "proxy variables" (hobbies, career gaps) to penalize female candidates, even with gender markers removed.

  • AI drugs hit trials: Several AI-discovered drugs reached mid-to-late stage trials in 2026, testing whether ML can navigate biology and regulatory approval.

  • Regulation battle: Trump's executive order targets state AI laws. California, Texas, NY, and Illinois have their own regulations taking effect in 2026. Constitutional showdown incoming.

Thanks for reading. Our mission is to educate as many people as possible around AI literacy - see you next week.

- Drew & the rest of the humans behind The AI Digest

Keep Reading