It's over.

TL;DR

• 1

  Fable looked absurdly capable in quick prototyping: Mo says he built 3 to 4 game concepts in one Friday, including Factorio-style browser games with 5 or 6 rounds of non-trivial rewrites, and the model "did a good job" throughout.

• 2

  Anthropic's position looks dominant by every metric he cites: He claims Anthropic has led the last 2 to 3 model cycles, is valued above OpenAI, has the talent edge, and holds roughly 90% of the enterprise API market versus OpenAI's 10%.

• 3

  He does not buy the official safety framing: Mo's read is that the Fable takedown drama was about power, capital, and government influence, not a genuine fear that the model finding bugs in code represented some new catastrophic capability.

• 4

  The White House-Anthropic clash reads like politics, not process: He points to Politico's reporting on the alleged "wellness retreat," the 90-minute deadline, and prior tension after Anthropic reportedly denied Pentagon requests for broad model access, including weapons-related use.

• 5

  Executives get "one-shotted" by demos because they judge outputs, not code: Mo says prototype sessions can feel magical if you never inspect implementation, echoing Andrej Karpathy's line that looking at the code gives you a heart attack.

• 6

  His bottom line on AI labor is more sober than replacement hype: The model still needed him prompting nonstop from 8 a.m. to 3 p.m., so his view is that AI is powerful but fundamentally an inanimate tool that augments humans more than it replaces them.