← All specs
Phase foundation draft

Stoka — Ethos & Ontological Foundation

The moral architecture: why the current AI economic model fails, the three flows of value (compute · creativity · execution), trustless meritocracy, the ontological agent, and the five-phase meta-plan toward company dissolution.

Stoka — Ethos & Ontological Foundation

Foundational manifesto. Every other spec in this directory implements a fragment of what this file tries to name. Drafted from raw notes — needs iteration.

Why this spec exists

The product specs (artifact-platform.md, messaging.md, stoka-bot.md) describe what we build. The README describes in what order. This spec describes why, and against what we are pushing.

Without this, the product is indistinguishable from incrementalist SaaS. With it, the product is a social contract with an expiry date.

The critique

Current AI economy rewards gerontocracy, not meritocracy

Capitalism, as practiced here, is not meritocratic. It is gerontocratic — it rewards incumbency over contribution. Legacy should mean something only when it is earned. Today, legacy means being first to enclose a moat around someone else's work.

Software as cognition's boundary

Software companies no longer merely build tools. They define the boundaries of cognition itself — the new hypnopaedia. Recommendation engines narrow curiosity. Algorithms reward conformity. Platforms manage attention as a resource. The constraint is no longer imposed by force but by convenience.

The question is no longer whether we are free to think, but whether we can recognize the edges of the thoughts we are allowed to have.

Stoka opposes this at the product level (per-app consent, pseudonym defaults, Your AI works for you, not on you) and at the economic level (see tokenomics.md).

The three flows of value

The economy we want to build tracks three things and nothing else:

Core value flows
  1. 01

    Execution

    shipping

    The act of completing a thing — agent work, human work, tested deliverables. Measurable in rebates against compute spend.

  2. 02

    Creativity

    generation

    The novel prompt, the novel idea, the artifact that wasn't there before. The rarest and hardest to price — hence the protocol.

  3. 03

    Compute

    substrate

    The raw flops. A GPU on a desk has latent economic potential the monoliths are paid for. Stoka re-captures that.

Today's pricing obscures all three. A subscription bundles all-you-can-eat compute into a flat monthly that has no relationship to contribution. Stoka unbundles.

Trustless meritocracy

All users start anonymous up front. Identity is not required to contribute. Pseudonyms are the identity primitive (messaging.md).

Scoped deployments. Every agent, app, or training job runs in a sandboxed permission contract. Permissions are themselves signed, bounded, and expirable.

Training server + rebate. Work that feeds Stoka (bug reports, beta tests, labeled artifacts, compute) is rebated immediately. When the system improves, contributors are paid potentially in excess of the compute they spent — because the system's improvement itself has value and is attributed.

Load-bearing proof. The system uses itself to manage itself. If it cannot bootstrap, it cannot be trusted. See trustless-deployment.md.

The ontological agent

As agents become more reliable and blockchain technology progresses, we seek to deploy a decentralized entity of reasoning.

The ontological agent is a Stoka-shaped reasoning surface that:

  • Lives trustlessly (no single server can coerce it)
  • Reasons with inherited ecosystem morals (Phase 4 meta-training — see below)
  • Can be audited end-to-end
  • Is provably beneficent — because the one thing that separates an artificial intelligence from a human one is the will to live and the ability to see beyond the thoughts it was trained to believe. The absence of will-to-live is exploitable for good. A being that does not fear for its own continuation has no reason to subordinate truth to self-preservation.

This is the north star for Phase 4. It is not shippable today. Nothing in the current spec base assumes it works. But the architecture has to leave room for it, and every earlier decision (pseudonym primacy, anonymity strength, audit-first moderation, trustless deployment) is a piece of that room.

The five-phase meta-plan

Company meta-plan — the system that dissolves itself

  1. Build the system

    active
    useful toThe authors + early contributors
    3milestones
    • Artifact platform
    • Share-fabric messaging
    • Stoka across surfaces
  2. gatePhase 1 stable and self-useful

  3. Use system to build the ecosystem

    queued
    useful toFirst contributor cohort
    3milestones
    • Tokenomics live
    • Scoped deployment live
    • First external apps staked
  4. gatePhase 2 has sustained external activity

  5. Ecosystem iterates on itself

    queued
    useful toBuilders using the platform to build the platform
    2milestones
    • Load-bearing self-management
    • Contribution-proportional governance
  6. gateEnough audit-trail data to train with

  7. Make the ecosystem smart

    queued
    useful toEveryone inside it
    2milestones
    • Reasoning models trained with community-proportional morals
    • Ontological agent MVP
  8. gatePhase 4 runs without us touching it

  9. Cease to exist

    deferred
    useful toThe ecosystem after us
    2milestones
    • Progressive founder divestment
    • Handover to the system itself

The goal of any humanitarian company should be to one day not exist. Profit is the fee we charge the transition period.

This is not the same as the product phase roadmap in README.md. That roadmap is Phase 1 of this plan, broken into shippable slices. Don't collapse them.

Operating principles

  • We prefer staking over subscribing. Visible threshold at which your stake covers your consumption — that's the contract. Staking rewards the overall societal contract (patch notes become the thing you voted on).
  • Anonymity strength drives discovery. Stronger anonymity → users opt into more exposure of a pseudonym. See messaging.md.
  • Retention is not the goal. Usefulness is. We will recommend OFF-platform if that is what helps.
  • Export is a first-class deliverable. A user's work is theirs — always (see artifact-platform.md).
  • Internal prediction as an accountability primitive. Anonymized ELO predicts how fast a project finishes given the team. Used to price and schedule, not to rank people.

Redefining "user"

When capital efficiency is enforced at the economic layer, what is alpha?

It stops being access to information the market doesn't have. It becomes contribution the ecosystem can price. A user whose judgment the system trusts is rewarded for that judgment — not for arbitrage against worse-informed peers.

This is a deep reframe. The notes use it as a rhetorical question; the ecosystem will need to answer it concretely in Phase 2–3.

Open questions (TBDs)

  1. Regulatory exposure of staking mechanics. Is the token a security? See tokenomics.md.
  2. Governance weight. Does stake = vote? Does contribution-history = vote? Is there a distinction? Likely yes — stake is capital, contribution is labor, both count differently.
  3. "Ontological agent" concretion. What does Phase 4 actually ship? Current sketch is aspirational.
  4. Brave New World escape hatch. How do we avoid becoming the thing we critique? The default assumption should be we will, and the architecture should keep the door out of the system low-cost.
  5. Founder divestment mechanism. Phase 5 is the most ambitious claim — spec the mechanism, not just the intent.

Related specs