S

self-improving

A predictions ledger with forced settlement for AI agents: log every judgment as a bet, settle it against reality, distill recurring biases into rules. Learns from outcomes, not corrections.

A skill and toolkit for running self-improving AI agent teams. Codex or Claude Code orchestrates over Symphony workers with Linear-managed execution. Includes dispatch, review gates, concurrent workers, runbook-driven learning loops, and durable guidance promotion. No auto-merge.

[2025] Autonomous Claude Code plugin with pattern learning and skill auto-selection. Features 27 specialized agents across 4 collaborative groups, quality control automation, and intelligent task execution without human approval at each step.