The vNext Roadmap
FLUX 0.3.x is a rigorous core. The most advanced private implementations of this idea are already ahead of it in places — richer population models, versioned module reuse, runtime policy enforcement, graded calibration. The roadmap's bet:
Make the open standard a strict superset. An advanced implementation should migrate because it loses nothing it values and gains everything the standard offers — migration as an upgrade, never a compromise. Where an implementation breaks the spec today, it is usually ahead of spec: the right move is to pull those ideas upstream, not sand them off.
The full proposal is nine RFCs — see The Nine RFCs — landing across staged releases, governed by five principles.
Design principles
| # | Principle | Meaning |
|---|---|---|
| P1 | Additive, not breaking | Every 0.3.0 document stays valid in every 0.3.x/0.4.x release. New power is opt-in. |
| P2 | Closed core, open edges | additionalProperties: false stays; the extension port is the one sanctioned escape hatch. |
| P3 | Profiles, not one size | Core / Enterprise / Runtime conformance — a validator, a governed org, and a live engine each get an honest bar. |
| P4 | Executable, not just checkable | The spec grows a runtime enforcement contract, not only a schema. A standard you can run beats one you can only lint. |
| P5 | Provenance everywhere | Documents and seam crossings carry verifiable provenance, so "what was proven" is auditable end-to-end. |
Staged releases
| Release | Scope | Status |
|---|---|---|
| 0.4.0 | RFC-01 trait distributions · RFC-02 extension port · RFC-03 module supply chain (flux.lock) · RFC-07 semantic-model port · RFC-08 UI hints — all additive, zero breakage | ✅ shipped |
| 0.4.1 | RFC-05 credibility scorecard · RFC-06 seam version ranges + provenance — additive | ✅ shipped |
| 0.5.0 | RFC-04 runtime enforcement contract + conformance vectors · RFC-09 deterministic signed evidence bundles | ✅ shipped |
Each release is independently adoptable, and each RFC lands the repo's way: one PR touching schema + schema-diffs + examples + tests together.
The migration ledger
The decision rule for any advanced implementation: migrate when this table has zero losses. Under the full roadmap:
| What an advanced implementation relies on | Under 0.3.0 | Under the roadmap |
|---|---|---|
| Trait distributions per identity | lost — flattened to scalars | kept — RFC-01 typed distributions (shipped) |
| Dialect vocabulary, audit/retention fields | illegal — forces a fork | kept — RFC-02 extension port (shipped) |
| Versioned modules & cross-bundle reuse | lost — flat in-bundle ids | upgraded — RFC-03 semver refs + lockfile (shipped) |
| Runtime policy enforcement | dropped — validator only | upgraded — RFC-04 enforcement contract (shipped) |
| Closed-loop calibration verdicts | partial — drift only | upgraded — RFC-05 scorecard (shipped) |
| Contracts on older FLUID versions | stranded — frozen enum | upgraded — RFC-06 negotiated ranges (shipped) |
| Semantic metric binding | opaque string | kept — RFC-07 semantic-model port (shipped) |
| Form-based authoring tools | non-standard | kept — RFC-08 UI hints (shipped) |
| Evidence packs | local pass/fail only | upgraded — RFC-09 signed bundles (shipped) |
Zero losses, four keeps, five upgrades — all nine now shipped. A private dialect is, at this point, just maintenance debt: everything it needed is in the standard, and the standard's top tier is the cheaper place to live.