An order, without dates.
LIVEX's roadmap states an order of work, never a calendar. Only what is instrumented by a milestone, a decision or a test is included — vague intentions stay out.
The order, not the promise.
The phases follow one another in a mandatory sequence, but their calendar depends on availability and on intermediate validations. This choice preserves rigour: an order kept is better than a date not kept. Publication milestones are always downstream of acceptance, never the reverse.
A useful roadmap is recognisable by the fact that one can lose points on it. This one loses three, stated below: a milestone delivered but not crossed, a performance gap not closed, a calibration not done.
Delivered, in reserve, to do.
Delivered
Present in the code and covered by tests. A verifiable achievement.
Delivered but not activated
The mechanism exists and works, but it is disabled by default, not calibrated or without observable effect. A capability in reserve, not a scientific achievement.
To do
Identified, ordered, not started or not crossed.
Confusing the first two categories is the most costly mistake of such a project: a mechanism delivered but inert produces nothing, and presenting it as delivered would lead one to believe that the world contains what it does not yet contain.
From foundation to consolidation.
| Phase | Title | Objective |
|---|---|---|
| 0 | Foundation & governance | Repository, conventions, pipeline, licence, governance. |
| 1 | General framing | Vision, architecture, inter-component contracts, glossary. |
| 2 | Engine (SYNE) | Heart of the simulated world: cognition, action, internal observation. |
| 3 | Observation (ECHOS) | Analysis, metrics and control of experiments. |
| 4 | Unreal integration (PRISM) | Unreal project and PRISM-LDK plugin. |
| 5 | Consolidation | Consistency of contracts, FAQ, contribution, cross-review. |
The phases follow one another without exception: the next one starts only once the milestones of the current one are crossed.
What is crossed, what remains open.
| Workstream | Status |
|---|---|
| SYNE engine | Twelve technical phases delivered (foundation, BDI and perception, memory and beliefs, decision, actions, communication, groups, resources, loyalties, observability, performance, tests and persistence). Capabilities in reserve: seasons, territories, constructions, inventory, commitments. |
| ECHOS observatory | Nine phases delivered, two withdrawn (integrated interface and Electron shell, ADR-007 of 5 October 2026). Main remaining workstream: the dynamic viability index. |
| PRISM rendering | Working version. Transport contracts stabilised; validation against real SYNE pending. |
| Launcher | Milestones G0–G4 largely present. G5 (complete session with real ECHOS): delivered on the Launcher side but not crossed. G7 (Immersion mode): locked, for lack of a compliant PRISM. |
| Installer | At the specification stage, without a line of code. Five milestones defined; Linux first. |
These statuses describe the repository as measured on the date they were recorded, not a calendar: no dated milestone is set before the phases in progress are consolidated.
What the project forbids itself.
Setting these avenues aside is not a judgement on their interest: it is a refusal to inflate the target.
Multi-machine
Distribution across several machines is documented but not implemented, and does not count as work.
Other systems
Only Windows and Linux are targeted; other platforms are out of scope.
Language models at the heart of the world
Recourse to a language model as a decision engine remains a documented avenue, not a commitment.
Delivery dates
No dated milestone is set before consolidation of the phases in progress.