Working with large changes

IntentLang is useful for one mission. It becomes essential when there are many.

A single day of Claude Code or Codex can generate dozens of .intent files. A release can touch hundreds. Reviewing them one file at a time does not scale, and it hides the questions that actually matter: what changed, what is risky, what is verified, and what blocks deployment.

This is the hub for the concepts IntentLang uses to answer those questions across many missions. Start with the Mission Atlas, which maps what exists; each concept below has its own page, and the tutorial walks all of them end to end on one example.

Five teaching lines carry the whole model:

• Mission Atlas shows what exists.
• Build Session Digest shows what changed.
• Proof Matrix shows what is trusted.
• Risk Radar shows what to review first.
• MVP Readiness shows whether it can ship.

The concepts

Mission Chain

A connected sequence of missions that forms one user or system flow. Missions measure per-unit correctness; chains measure end-to-end readiness. A release with 200 green missions can still ship a broken signup if one link is unverified. Full page: Mission Chains.

Build Session Digest

A summary of everything that changed during one work session, expressed in intent rather than files. It is what you read after an agent has been working. Full page: Build Session Digest.

Semantic Diff

A diff by meaning, not by file: guarantees added, never rules weakened, proof gone stale. It surfaces the changes that change trust. Full page: Semantic Diff.

Proof Matrix

A table of verification status across many missions, so a reviewer can scan a whole release quickly instead of opening each file. Full page: Proof Matrix.

Risk Radar

A priority list of the missions that deserve review first, so a team does not review 200 missions as if they were equal. Full page: Risk Radar.

MVP Readiness

A classifier (demo_safe .. production_ready) that lets teams ship fast without pretending everything is production-ready. Full page: MVP Readiness.

AI-generated missions

The review model for volume: how IntentLang keeps dozens of agent-authored missions reviewable through provenance, risk, and proof. Full page: AI-generated missions.

Release Story

A Release Story is a human-readable summary of what an MVP or release can do, and what is not yet verified. It is not marketing copy. It is a trust-aware release narrative: every capability it claims is tied to the missions and proof behind it, and it states plainly what is unverified. It is the output you can hand to a stakeholder.

See a worked example at examples/mvp-customer-portal/release-story.md, produced (in the planned model) by intent release ./intent --mvp alongside the MVP Readiness report.

Tutorial

From 200 missions to one Release Story walks the whole path on the customer portal example: index, atlas, chains, digest, proof matrix, risk radar, semantic diff, readiness, and the release story.

Command status

This repo teaches the concepts and ships the example fixtures. The first aggregation command, intent index, is shipped; the rest are owned by the SkillsTech Compiler and documented here as planned until they ship. The single-mission commands that already exist (intent check, graph, proof, build, lift, approve, drift, handoff) are covered in the compiler contract.

Nothing here claims production readiness. These views report what the repo can prove and what it cannot. Verifying the running code against the missions is OpenThunder's job.