Lucid Lake Support

Blog: Snoot

  • Every claim on a clock: how AI agents made our explainer films #snoot

    AI agents made five short explainer films for Snoot, each claim tied to a measured run and gated by code. In the 42 s pilot, all 86 spoken words matched the script, and the quietest word sat 16.9 dB over the music.

  • What a headless browser caught that our eyes missed #snoot

    Our page checker opens every page in a headless browser and judges contrast, colour-blind contrast, keyboard order and layout. It caught a heading that passed AA for normal vision at 4.54:1 but fell to 3.85:1 in a red-blind view.

  • Engineering a plush robot before it exists #snoot

    Snoot has no physical prototype yet, so every motion clip must pass servo limits, a torque model and a skin-clearance check in simulation before we render it. In a re-run on 13 Nov all 33 clips pass, the worst at 0.80 of a 0.9 servo line.

  • One eye, three renderers #snoot

    Snoot's screen eyes are drawn three ways, in Python, float C and fixed-point C, and a test holds them together. All three agree to within 1 ΔE on 99 % of the eye; the fixed-point build is correct but still too slow for the chip.

Aptmark is private

Enter the password, or ask me for access.

Ask for access