Determinism for AI Coding ?
No spec. No trace. No control.
Without Governance
Undisciplined AI coding
- random outputs
- unpredictable results
- specification drift
No clear path
Determinism for AI Coding
Governance over chaos.
With Determinism
Governed AI development
- predictable outcomes
- enforced phases
- spec-to-code traceability
No shortcuts. No going back.
Core Concept
Phase Discipline
No implementation without specification. Phases are enforced, not suggested.
No shortcuts. No going back.
RFC is Law
Specifications are the source of truth. Code that conflicts is a bug.
Gates are Real
Each phase has explicit requirements. No shortcuts, no exceptions.
Work is Traceable
Every task links back to the spec that authorized it. Full audit trail.
Ecosystem
The Suite
A complete toolkit for governed AI development.
govctl
LiveOpinionated governance CLI for RFC-driven AI-assisted software development. Enforce phase discipline on every feature.
skillc
LiveDevelopment kit for Agent Skills. Author, validate, optimize, and instrument skills for AI agents.
jjgov
Coming SoonMulti-agent collaboration workflows built on jujutsu. Coordinate multiple AI agents with version control.
everevolve
Coming SoonProject rules generated from commit history. Learn governance patterns from how your team actually works.
The Problem
AI coding is powerful.
But undisciplined.
Fast typing, slow thinking. Without governance, AI amplifies chaos.
Without govctl
Chaos timeline
- 1 Day 1: "Let's add caching!"
- 2 Day 2: AI generates 500 lines of Redis integration
- 3 Day 7: "Wait, did we agree on Redis or Memcached?"
- 4 Day 14: Half the team implements one, half the other
- 5 Day 30: Two incompatible caching layers, no spec, nobody knows why
With govctl
Deterministic timeline
- Day 1: govctl rfc new "Caching Strategy"
- Day 2: RFC-0015 defines: Redis, TTL policy, invalidation rules
- Day 3: govctl rfc advance RFC-0015 impl
- Day 7: Implementation complete, traceable to spec
- Day 14: govctl rfc advance RFC-0015 stable
In Action