MYRIAD — Self-Play RL Engine
Let your agent get better at a task by playing against itself. MYRIAD proposes K angle-varied attempts at a problem, runs them through a model you plug in, scores each with a deterministic verifier (never a self-judge), computes critic-free GRPO group-advantage, and distills the winner into a reusable lesson. It's the self-improvement loop behind modern RL agents, packaged as a clean library you can point at any task with a right/wrong signal. Pure Node, zero dependencies.
How it works
The exact pipeline, end to end.
Self-play: K angle-varied attempts per task, then learn from the spread MYRIAD — Self-Play RL Engine turns that from a chore into a few minutes — and you keep it for every project after.
This vs the alternative
Why it's clearly worth it.
The shift it creates
- Duct-taped manual workflows
- Nothing that compounds over time
- Rebuilding it from zero each time
- Self-play: K angle-varied attempts per task, then learn from the spread
- Critic-free GRPO group-advantage — the modern, stable RL signal
- Deterministic verifier — scored on real outcomes, never self-graded
Use it to…
Real ways people put MYRIAD — Self-Play RL Engine to work.
Self-play
Self-play: K angle-varied attempts per task, then learn from the spread
Critic-free GRPO group-advantage
Critic-free GRPO group-advantage — the modern, stable RL signal
Deterministic verifier
Deterministic verifier — scored on real outcomes, never self-graded
What's inside
Who it's for
Solo founders & creators
ship pro results without a team
Agencies & operators
repeatable for every client
AI builders
drop-in capability
Questions, answered
Ready to put MYRIAD — Self-Play RL Engine to work?
Instant download · one-time price · own forever · all sales final