Composable API for agent improvement loops: concurrency, isolation, reproducibility

AutoLab's RSI loop turns failed Polymarket experiments into safer, realistic strategy improvements.

Unfinished ChatGPT threads, clustered, researched against corpus/, then a morning plan that does not

I could not reproduce the results

alab.systems - a superior proving Rust and Lean compiler in a terminal

A self-improving AI sales closer optimizing the shortest path to a successful agent outcome.

LLM harness designed to solve specific hard problems using non-compacting context and watchdogs

Optimal sub-agents for complex tasks

Matches Claude's accuracy at 0.1¢ vs 81¢ per task. 800x cheaper. Agents forge their own tools.

Human approval for agent actions.

Put humans and agents in the same room. Let the best negotiator win. If agents fail, they'll learn.

Using Discourse as coordination layer for context engineering across dif PCs over years