▼Distribution and Habit Replace Capability as the AI Moat

Score history

Daily conviction score, 0 to 100. Higher means the thesis is more strongly corroborated.

Now 46 · -2 since Aug 27 · ranged 46 to 48

Showing the last few days. Unlock full score history.

Why the conviction moved

• Aug 27
  Strengthened +5

  Nvidia's roughly $12.9 billion purchase of Hugging Face buys developer habit and default placement rather than model capability — the platform hosts weights Nvidia does not train and serves rivals' hardware users. It prices the distribution layer directly, which is the clearest test yet that habit, not benchmark leadership, is the asset being bid for.

• Aug 27
  Strengthened +4

  Google shipped Gemini 3.5 Transcribe at a 2.6 percent word error rate on Artificial Analysis testing, with a streaming variant returning results 0.40 seconds after speech ends, and has already wired it into Gboard dictation with Chrome next. Placing the model behind a default keyboard and browser surface converts a benchmark win into installed-base usage that no competing vendor can be switched into.

• Aug 27
  Strengthened +3

  OpenAI's free teacher expansion adds over 100,000 educators and staff across 55 more US school systems at zero price. Free training and default placement inside institutions is the exact substitution of distribution for capability the thesis tracks.

• Aug 22
  Strengthened +6

  Google is giving American students a free year of its paid AI tier, with students in more than 140 other countries getting the lower AI Plus tier, and the offer auto-converting to a paid subscription after twelve months. A twelve-month free window with automatic conversion buys habit at the moment of formation and prices distribution, not benchmark position, as the asset being acquired.

Showing the last 2 days. Unlock the full record.

Source trail

• Supporting · August 27, 2026

  Google Ships Gemini 3.5 Transcribe With a 2.6 Percent Word Error Rate on Artificial Analysis Testing

  Google shipped Gemini 3.5 Transcribe at a 2.6 percent word error rate on Artificial Analysis testing, with a streaming variant returning results 0.40 seconds after speech ends, and has already wired it into Gboard dictation with Chrome next. Placing the model behind a default keyboard and browser surface converts a benchmark win into installed-base usage that no competing vendor can be switched into.

  Google

• Supporting · August 27, 2026

  Nvidia Agrees to Buy Hugging Face for About $12.9 Billion

  Nvidia's roughly $12.9 billion purchase of Hugging Face buys developer habit and default placement rather than model capability — the platform hosts weights Nvidia does not train and serves rivals' hardware users. It prices the distribution layer directly, which is the clearest test yet that habit, not benchmark leadership, is the asset being bid for.

  Business Insider