Motion Capture Lab
ResearchMotion capture studio on Floor 10 for embodied-agent training, performance capture, and character-animation datasets.
Four builds, two levels of done. The catalog is the source of truth; add a project here when it goes external.
Motion capture studio on Floor 10 for embodied-agent training, performance capture, and character-animation datasets.
Fully agentic wetlab on the floor — AI agents run pipetting protocols and biology workflows on Opentrons hardware via MCP.
Spatially-alive agentic floor — members interact with FT10 without any personal hardware. Six frontiers: acoustic localization, RF sensing, vision, voice, member intel, directional audio.
AI-agent security audits + honeypot infrastructure for the MCP era. Verified-defender positioning targeting Anthropic CVP and OpenAI TAC tiers.
An in-person benchmark for AI-built games: six models build the same arcade game from one strict brief, and humans play the versions blind on a physical cabinet and vote. The aggregate human vote — not an automated metric — is the model's score. 3 briefs × 6 models × 18 named games.
Living Portraits: two LED panels on the floor holding self-aware AI portraits with a memory loop. A slow local model gives each character a journal it reads back as memory, and lets it imagine new poses for itself. Every image made with Midjourney.