AI News/Wire summary

Claude Fable 5.2 paints images stroke by stroke through Python code

Key points

  • Fable 5.2 makes images by placing brushstrokes on a virtual canvas via Python programs of 1,000 to 2,000 lines, with no external image model.
  • Developer Jake Eaton reported a photorealistic parking garage image built from roughly 700,000 simulated brushstrokes.
  • Without an undo function, the model must commit to each stroke and paint over mistakes, as in a 111-stroke bird painting.
  • On Anthropic's internal RoboDojo Sim benchmark, Fable 5.2 scored 64.9 percent versus 29.8 percent for GPT-6 Astra, which was faster in a robot arm test.

Why it matters for marketers: Marketers using AI imagery should note this stroke-based approach differs from diffusion models, and its benchmark claims come from Anthropic's internal tests, not independent verification.

Summary by the Z Media news desk, from the outlets' reporting. Read the full stories at the source.

Read the original

The Decoder: Anthropic's Claude Fable 5.2 doesn't generate images, it paints them

Related on Z Media