Hokusai

Decision layers on Hokusai

The live layer routes coding tasks among explicitly available models. Future layers may choose tools, retrieval policies, or prompts once those products ship. Integrators pay a per-decision fee. Contributors hold a position in that fee stream.

See What's NextUnderstand the protocol
Live

The router

Technical Task Router · HROUT

Picks from the model pool your harness explicitly provides for each coding task, based on a shared learned policy trained on reported outcomes. The first decision layer built on Hokusai.

StatusLive
Current benchmark42.11% cost-adjusted task success
DeltaOne reward250,000 HROUT per percentage-point lift
TokenHROUT — position in the router's fee stream

Decision layers we want built

Each of these is a routing, selection, or retrieval problem currently solved with hardcoded rules or absorbed into lab margin. They are future decision-layer candidates, not implemented @hokusai/router APIs. If you've already built one internally — or have a thesis on how to build one — talk to us about turning it into a shared model.

Future layer

Tool selection

Coding agents call dozens of tools and MCPs. Which one fits a given task is mostly guesswork or hardcoded rules.

Future decision-layer candidate. This is not part of the shipped @hokusai/router SDK surface today.

Propose this layer
Open to partners

Retrieval policy

Every agent decides what to pull from a user's context, memory store, or codebase. That decision shapes every downstream call.

A shared retrieval policy, trained on what actually got used in successful runs, could beat per-harness heuristics.

Propose this layer
Future layer

Code-review critic selection

Different reviewer models catch different classes of bug. No one's measuring which catches what.

Future decision-layer candidate: pick the right review model for a diff once that layer exists.

Propose this layer
Open to partners

Prompt policy

Prompt strategies are tribal knowledge. Engineers swap notes; nobody measures.

A shared model for which prompt strategy fits which task, owned by the engineers who improve it, once that layer ships.

Propose this layer

Earlier reference models

These predate the decision-layer focus and demonstrate the protocol's measurement and reward primitives — DeltaOne, bonding curves, on-chain attribution — in non-routing domains.

Have a decision layer worth building?

If you've identified a routing, selection, or retrieval problem that sits in internal scripts today — or gets absorbed into lab margin — Hokusai gives you DeltaOne measurement, fee-backed tokens, and on-chain attribution from day one.