← Tomek Kaszyński

Everything I've built

16 projects, strongest first. I design the architecture and the pass/fail tests, and coding agents I direct write most of the code. Every number here has a receipt.

  1. Sep 2026TractatusA loop that plays ARC-AGI-3, puzzle games with no instructions. My code lays out the options and GPT-6 Astra only picks, so every move has a receipt.183/183 levels
  2. Sep 2026The 6-cent playerA cheap open model plays whole ARC-AGI-3 games start to finish, cold, and a separate checker replays every move. Four full games so far.9/9 for 6.5¢
  3. Sep 2026Built for PersonaMy pitch to be Persona’s CTO: a phone you can talk to that answers live and hands you buttons that open your real Messages, Maps and Calendar, plus my three proudest builds.live demo
  4. Sep 2026LocusSmall language models I train from zero get a sleep-like night between days of reading. They forget about two-thirds less, at three sizes. The same compute without sleep does nothing.40+ runs in 3 days
  5. Sep 2026Memory that sleepsQwen3-0.6B writes new facts into a notebook as it reads, moves them into its weights at night, and the notebook gets wiped. It still knows them. Fine-tuning stays near chance.88–93% kept vs 28%
  6. Sep 2026OnceA chat assistant on my Mac. Show it a task one time and it saves it as a small tested program, so next time it runs without calling the model at all.0 model calls on reruns
  7. Dec 2025TrendvoCo-founder, leading tech. I built the core end to end: traders’ track records verified straight from their broker.team of 9
  8. Mar 2026Emergent compositional communicationSolo arXiv paper. AI agents that see through different vision models invent a tiny shared language for hidden physics like mass and friction.80/80 seeds
  9. Aug 2026PRISMMy first ARC-AGI-3 player: plain code, no AI model. It cleared every public level in a minute, but only because I tuned it on those games. That’s why Tractatus exists.183/183 in 60 s
  10. Sep 2026CrucibleA laptop-only player that sees nothing but pixels. It cleared five puzzle games in under three minutes, using rules written for those five.41/41 in 174 s
  11. Sep 2026Learn once, replay freeA game’s rules, learned while playing, get saved. The saved model then plays the whole game again by itself.142/142 predictions, 0 calls
  12. May 2026ARC ForgeA machine that invents new puzzle games and checks that each one is solvable and actually new.2,027 of 8,030 kept
  13. Jun 2026ARC ScientistAn earlier puzzle agent run under strict referee rules. Held-out mini-games went from 4 of 12 to 9 of 12.9/12 held-out
  14. Sep 2026Persona, day oneWhen someone texts you, your Mac drafts a reply in your voice. Nothing sends until you tap, and every send leaves a receipt.$0.0001 a message
  15. Dec 2025SpecWiseA seatbelt for AI-written code: estimates how much of a repo AI wrote, flags risky files and can block unsafe merges in CI.
  16. May 2026Who said whatMeeting recordings to text with speaker labels, entirely on a Mac. No upload, no per-minute fees.