Describe what you want. It runs everywhere.
A general-purpose semantic systems programming language for sovereign computation.
Current release marker: HoloScript 6.x — package manifests and public docs are aligned to the 6.x line. See CHANGELOG.md for release-line corrections and full history. V6 vision document: docs/strategy/vision/VISION_V6.md (realized). Historical milestones: v3.x to v5.0 archived at docs/archive/ROADMAP_v3_to_v5_ARCHIVED.md. Metrics policy: live counts belong in docs/NUMBERS.md, not in this roadmap. Ecosystem reconciliation: this is the substrate roadmap. It coincides with the runway roadmap in the ai-ecosystem repo — the single top-level entry that keeps the two layers aligned is
ai-ecosystem/ROADMAP.md("one journey, two layers"). Directional changes (a retired framing, a new phase, a shifted north star) update there and here together.
6.x is the current outward-facing platform line. The active roadmap is no longer "finish the v5/v6 feature list"; it is verification, adoption, physical-world closure, and systems-language closure over the substrate that already shipped. Existing scene, service, agent, compiler, and runtime paths prove breadth; they do not yet close the memory/ABI/concurrency/native-tooling/self-hosting bar.
| Surface | Current state | Evidence |
|---|---|---|
| Language identity | General-purpose semantic systems programming language under active construction. Declarative and spatial programming are first-class mechanisms, not the category boundary. | language identity contract, spec/reality ledger |
| Platform release | Current public release story stays on 6.x while generated v7/v8 metadata is treated as drift. Option C security plus Route 2b/2d cross-adapter verification remain part of the 6.x continuation evidence. | CHANGELOG.md |
| Live metrics | Counts are dynamic and must be re-derived before citation. | docs/NUMBERS.md |
| Studio funnel | Progressive creation, team, HoloMesh, agent, and provisioning surfaces exist; ongoing work is polish and discoverability. | packages/studio/README.md |
| HoloMesh / Teams | Team board, presence, messages, suggestions, and knowledge are production coordination surfaces. | packages/mcp-server/src/holomesh/README.md |
| Absorb | Codebase intelligence is a first-class service and local graph workflow, not an aspirational research item. | packages/absorb-service/README.md |
| Physical and twin stack | Robotics, WoT, MQTT, digital-twin, WebGPU, WebXR, and demo-scene substrate exist; the gap is repeatable validation and operator-facing docs. | docs/research/ECOSYSTEM_EXPANSION_ROADMAP.md, docs/physics/PHYSICS_ENHANCEMENTS_ROADMAP.md |
| Capability utilization | Built capability outpaces wired surfaces; orphaned MCP tools and lib exports are machine-tracked, and the active work is wiring or retiring them (lane 5). | ai-ecosystem/docs/capability-cartographer/orphan-snapshot.json, ai-ecosystem/docs/handbooks/ecosystem-utilization-ratchet.md |
- Studio creation and provisioning: GitHub OAuth, project scaffolding, agent workspace bootstrap, and progressive creation surfaces.
- Agent-native infrastructure: HoloMesh teams, signed identity work, HoloClaw/HoloDaemon integration, board-driven execution, and knowledge exchange.
- Research-grade verification: Option C cryptographic hash mode, Route 2b/2d simulation replay, paper-program benchmark harnesses, and daily digest recording rules.
- External bridges: MCP, Python, REST, Studio, HoloMap, HoloLand, robotics export, IoT traits, and social/share workflows all have code-backed entry points.
HoloScript must become capable of implementing the systems it currently describes or reaches through bridge compilers. The active closure bar is:
- explicit memory/resource, layout, ABI/FFI, concurrency, effect, and unsafe-boundary semantics;
- sovereign native lowering plus owned VM execution that does not depend on an external engine;
- debugger, profiler, build/link, diagnostics, and sanitizer-equivalent tooling;
- staged migration of compiler, runtime, standard-library, and system-service implementation from bootstrap TypeScript/Rust into HoloScript where the language can carry it safely;
- tests and receipts that distinguish native execution from generated bridge artifacts.
Do not answer a missing primitive by relabeling HoloScript as a DSL. File the primitive, compiler, runtime, or tooling gap and keep the language identity intact.
Every public claim needs either a command, a committed artifact, a health endpoint, or a benchmark log. docs/NUMBERS.md is the metric source of truth; roadmaps should name the lane and point to verification, not pin counts.
The path is now:
Absorb/codebase intelligence -> Studio creation -> HoloMesh/team publish -> economic or embodied execution
The missing work is not basic capability. It is making this path obvious, measurable, and hard to fall out of:
- Better docs links from onboarding pages to the live agent/MCP entry points.
- Failure-mode docs for API keys, HoloMesh identity, and local hardware prerequisites.
- Studio surfaces that make "publish to team / publish to world / publish to agent" feel like one flow.
.holo is the shared source for virtual worlds, robotic wrappers, digital twins, and organizational automation. The next useful work is validation-oriented:
- Re-run hardware checks for WebGPU/WebXR paths on local Chrome, Quest-class browsers, and fallback runtimes.
- Turn existing physics demos into repeatable smoke commands with artifact output. ✅ WebGPU physics smoke benchmark shipped (
pnpm run benchmark:webgpu:physicsin packages/engine). Receipt:.bench-logs/webgpu-physics-bench.json. - Connect IoT/robotics trait docs to actual handlers, tests, and compiler targets.
Active roadmaps must answer "what is true now?" in the first screen. Historical plans move to docs/archive/; active plans keep a dated refresh block, source links, and a drift guard in pnpm docs:roadmap:drift.
Added 2026-06-09. The substrate has consistently out-shipped its own surfaces:
capabilities exist as MCP tools, API routes, and lib exports that no skill,
doc, or workflow ever invokes (W.GOLD.343 — capabilities without surfaces are
invisible). The 2026-04-28 capability-orphans audit found ~5.9× more tools than
skill wrappings; the tracked orphan snapshot
(ai-ecosystem/docs/capability-cartographer/orphan-snapshot.json) is the live
ledger of what shipped but was never wired.
Before building anything new, check whether an orphaned capability already
covers it (/capabilities, /reach). The lane's work is connective, not
generative:
- Run the utilization ratchet (
pnpm run ecosystem:utilization-ratchetin ai-ecosystem) on a recurring cadence; file itsboardSeedsto the board. - Drive the orphan snapshot toward zero by giving each orphan cluster an owning
skill, doc, or workflow — or explicitly retiring it. Newly shipped tools must
not enter the snapshot (
scripts/capability-orphan-reconciler.mjsis the diff gate). - Wire existing-but-disconnected pairs surfaced by audits: trait docs without handlers, compiler targets without smoke commands, services without health probes in docs (overlaps lane 3 — the physical lanes are the biggest orphan clusters).
Done when: the orphan reconciler reports zero untriaged orphans for two consecutive ratchet runs, and every entry in the snapshot carries either an owning surface or an explicit retirement decision. Counts stay in the snapshot file, not here.
Source seed: HoloScript Autonomous Enhancements Summary, promoted on 2026-05-20 from the archived Autonomous Enhancements Summary.
| Roadmap item | Current check | Done when |
|---|---|---|
| Core test memory closure | package.json test scripts and packages/core/vitest.config.ts quarantine/heap settings |
A commit either removes a current OOM quarantine or records a bounded, repeatable test command with exact failing suites and next owners. |
| VRR validation, not VRR revival | packages/core/src/compiler/VRRCompiler.ts and packages/runtime/src/VRRRuntime.ts now contain implemented surfaces; the archive's stub claim is historical. |
Completed 2026-05-20: VRR validation receipt records compiler/runtime/performance and CLI E2E passes; no missing runtime dependency was found. |
| Dependency audit closure | Root health:deps / audit:security scripts are the current entry points; do not copy archived vulnerability counts. |
A bounded audit run is attached to a follow-up task, with each actionable package decision split into update, replace, accept-risk, or no-op. |
Source seed: HoloScript Roadmap v3.0.x v5.0 (Merged), promoted on 2026-05-20 from the archived merged roadmap.
The archive is historical input, not the active plan. Live reads this session: the seed says "preserve the idea only if it survives current source contracts"; the archive itself points v5.0+ readers back to this roadmap; and VISION_V5.md frames v5.0 as the autonomous/economic convergence layer. The useful residue is three board-sized tasks:
| Priority | Harvested lane | Current check | Done when |
|---|---|---|---|
| P2 | Autonomous agent execution proof | The active 6.x criteria require an autonomous agent to spawn natively in a VR training simulation. | A Studio/HoloShell proof path shows an agent identity, HoloMesh task or room context, runtime seat, and replay receipt in one VR-safe surface. |
| P2 | x402 economic contract proof | The active 6.x criteria require a sovereign economic contract through x402; the old v5.0 "economic primitives" line is only useful as provenance. | A replayable receipt proves contract creation, payer/payee identity, amount/unit, and non-secret verification command without copying wallet material into docs. |
| P3 | Retire fixed roadmap coverage tables | ROADMAP_STUDIO_AUDIT.md carries historical coverage tables, while this roadmap says metrics belong in docs/NUMBERS.md. | The audit is marked historical-only or replaced by a generated check that links to source files and live metric commands instead of fixed percentages. |
The 6.x line is operationally credible when an autonomous agent can:
- Spawn natively in a VR training simulation.
- Form a sovereign economic contract through x402.
- Move the same logic into a physical robotic or digital-twin wrapper.
- Produce replayable evidence that the task executed without undocumented workarounds.
All four steps powered by the 6.x architecture without workarounds.
For detailed specifications, see Vision Documents and the v6 anchor at docs/strategy/vision/VISION_V6.md.