Skip to main content
MosEx Macchina Lab — ethical awareness (cloud, mind, heart)MosEx Macchina Lab

Transparency

Roadmap & collaboration

This page lives on its own so funders, partners, and contributors can bookmark it and see what already exists, what we plan to build next, and what we need from others — in one place. It complements the investors scope narrative without overloading that page.

Maturation disclaimer

Ethos Kernel is actively maturing: runtime, audit trails, hub/DAO scaffolding, and tests land first so we can grow toward honest labeling, calibrated weights, and governance experiments without rewriting the stack every month. Some names and mocks still run ahead of the numeric core — that gap is tracked openly.

Maturation & critique track

Seven consolidated items (two independent reviews; duplicates merged). Full issue bodies to paste into CRITIQUE_ROADMAP_ISSUES.md; create tickets on GitHub Issues.

  • P0 — Honest naming: `weighted_ethics_scorer` canonical; THEORY + ADR 0009 aligned (`bayesian_engine` shim).
  • P0 — Input trust (merged): MalAbs on chat **and** LLM perception JSON (GIGO); hardening + optional local classifier; one threat-model doc.
  • P1 — Empirical pilot: labeled scenarios / agreement methodology (not certification).
  • P1 — Core path map + pip-installable **core** spike (MalAbs→will vs narrative theater).
  • P2 — Heuristic ethics & HCI: poles honesty; weakness/PAD vs operational trust in critical profiles.
  • P2 — Governance: MockDAO exit criteria + **L0 immutable buffer** vs community drafts (honest politics).
  • P3 — `KERNEL_*` consolidation via profiles + deprecation policy.
  • P3 — Async LLM: cooperative cancel / `httpx` async (ADR 0002 remainder) after thread offload + turn timeout.

Current foundation (in the repo)

Shipped and inspectable today — not a future promise.

  • Python ethical kernel with multipolar poles, Bayesian-style scoring, MalAbs vetoes, narrative memory, forgiveness, Ψ-sleep (simulation), and optional LLM interface layer.
  • FastAPI WebSocket runtime (`python -m src.runtime`): chat, optional sensors (v8), multimodal/vitality/epistemic telemetry, V11 judicial escalation + V12 moral hub / mock DAO in snapshots (schema v3).
  • Persistence: JSON checkpoints, optional Fernet encryption; conduct-guide JSON on disconnect; reality verification (lighthouse KB); named runtime profiles for CI.
  • LAN + smartphone thin client (`mobile.html`, scripts), nomad bridge docs; extensive `pytest` suite on GitHub Actions.
  • Public Next.js site: landing (updated for runtime/nomad), theory math showcase, investors, interactive dashboard, one-pager — Apache-2.0.
  • Documentation: README, THEORY_AND_IMPLEMENTATION, RUNTIME_*, LOCAL_PC_AND_MOBILE_LAN, NOMAD_PC_SMARTPHONE_BRIDGE, SECURITY, bibliography.

Near-term implementation roadmap

Concrete technical directions — priorities may shift with funding and partners.

  • Maturation track (P0–P3): see “Maturation & critique track” above and docs/proposals/CRITIQUE_ROADMAP_ISSUES.md — consolidated backlog (perception + chat input trust, pip core spike, HCI/weakness, L0 vs DAO).
  • Richer test matrix: edge cases, adversarial prompts against the LLM boundary, regression suite for domain additions.
  • Reference “sensor → signals” adapters (documented stubs or one hardware-agnostic demo path) without claiming product certification.
  • Conversation/session state design: multi-turn loops that keep the kernel authoritative over policy.
  • Dashboard build pipeline: pre-bundle JSX to reduce in-browser Babel reliance (security & performance).
  • Identity & voice guardrails: structured profile the kernel owns; LLM restricted to phrasing.

Mid-term (aspirational)

Depends on governance, capital, and institutional alignment — listed for honesty, not as commitments with dates.

  • Governance beyond mock DAO: requirements, threat model, and optional testnet-style experiments — only with legal and partner clarity.
  • Pilot integrations with robotics or vehicle stacks in **staged, supervised** environments — never as a substitute for regulation.
  • External ethics or safety review cycles documented in the open.
  • **Architecture:** incremental hexagonal-style boundaries (ports & adapters) where it pays off — e.g. LLM backends, governance/DAO, and episode persistence — so the ethical core can swap implementations without rewriting the pipeline. Applied gradually, not as a big-bang refactor.

For mock vs on-chain governance narrative, see BlockChainDAO.

Funding & collaboration needs

Where outside help directly accelerates the roadmap.

  • Sponsored engineering

    Time to implement the near-term roadmap, maintain CI, and respond to security reports — ideally scoped grants or part-time contracts.

  • Research & ethics partners

    Universities, labs, or civil-society groups for co-authored scenarios, red-teaming, and policy-facing briefs.

  • Hardware / integration partners

    Short spikes to map real sensors and actuators to the kernel’s signal model — no implied warranty on deployments.

  • Operations

    Sustainable hosting, domain/DNS hygiene, and occasional legal review for disclosures and contributor agreements.

  • Community

    Contributors for docs, translations, simulation scenarios, and reproducible bug reports.

Discussion forum (planned)

Coming later — not a commitment with a launch date.

We plan to add a hosted discussion forum for open threads on pending implementation areas, roadmap trade-offs, and community proposals — complementary to the code repo. Until that exists, use the collaboration GitHub template and public GitHub Issues; enable Discussions on the repository if the maintainers turn that on for broader Q&A.

Engage

Use the partnership / funding GitHub template for scoped proposals. For security-sensitive reports, see SECURITY.md. Donation channels: Donate.

Last updated with the repository; substantive roadmap edits should land via PR or tracked issue.