Studies

  1. 01

    Persistent AR

    Beta · 2022–2026

    What would it take for anyone to build a spatial world on a phone — and find it exactly where they left it?

    CVPR 2026Workshop paper on persistent, editable 4D world models — second author
  2. 02

    Text-to-world generation

    Prototype · 2026

    Can a sentence compose a world to an art director’s standard?

    1,576Reusable components, catalogued from 120 display rooms
  3. 03

    Semantic memory

    Research · 2026

    What if a model compiled what it knows once — and every answer stayed tied to its evidence?

    97%36 of 37 held-out reasoning tasks, matching a frozen LLM baselineOne synthetic microworld. A zero-parameter baseline scores 31/37, so the learned margin is 31→36. Not evidence of general reasoning.
  4. 04

    Auditable financial AI

    Prototype · 2026

    What would an AI agent need before it could be trusted with a founder’s books and taxes?

    0 → 280Tests in three days, as an audit became an integrity core
  5. 05

    Latent Compute

    Research · 2026

    Does a small language model with context from the device it runs on outperform the same model without it?

    7 gatesSet before building — all seven at once is the bar
  6. 06

    Financial forecasting

    Research · 2026

    Can a small research loop tell a real trading or forecasting edge from a lucky one before real money finds out?

    9.9%An ensemble over a single model on Brier score, 45 markets — the market price still led

Design & motion

Notes

Smaller projects, without a full study yet.