AI is going to change how we think and how we work. The busywork goes to the machine. People keep the judgment, and finally have room to use it.

Only 267 of the 41,884 known near-Earth asteroids have had their makeup measured, and makeup is what decides what an asteroid is worth. The rest were guesses. The model predicts each one’s type from data we already have, which takes us from 267 classified to 838 and makes the list actually searchable. Value comes from size, mass and makeup. How hard each one is to reach comes from published numbers where they exist, and an estimate everywhere else.

Each dot shows how it was classified

Classified

267 → 838

Catalogue

41,884 asteroids

Stack

Python · FastAPI · React

Each agent gets its own job and its own copy of the code, so several can work at once without stepping on each other, and you can always see who did what. When one needs a decision, it stops and tells you, and you pick up right where it left off. I started it as my own take on an open-source agent tool. Now I use it every day.

Each agent works on its own. It asks instead of guessing

Agents

Own identity

Isolation

Own copy of the code

Stack

Tauri 2 · Rust · React

Medical and legal records keep their meaning in their structure: sections, defined terms, references to other pages. Cut them up by length and the AI gets pieces that don’t make sense on their own. So Lexx splits documents along their own sections and keeps each piece with what it refers to. Every answer comes back in a fixed format and points to the page it came from.

It splits on sections, not word counts

Splitting

By section

Output

Fixed format, cites the page

Stack

Next.js · Claude API · TypeScript