← All specs
living

Stoka — Platform Mission & Roadmap

The map: what Stoka is becoming, why, and in what order. Start here.

Stoka — Platform Mission & Roadmap

Thesis

Your AI work, remembered, retrievable, and — optionally — shareable.

Stoka is evolving from a portfolio site into a platform where artifacts (structured captures of AI conversations, prompts, specs, essays, codebases) are created, discovered, and — only if the author chooses — shared. The core unit is the artifact. The engine that curates, structures, and surfaces them is Stoka itself — a single editorial presence running across every surface of the product.

Three principles that everything hangs off

1. Solo-first, not bootstrap-solo

The product must feel like an amazing personal tool before anyone is invited. N=1 users get real value. A builder using AI every day should want to use Stoka whether or not it ever becomes social. Capture your conversations, build your library, retrieve with natural language — that alone is the product. Sharing is downstream.

2. Sharing, not posting

The center of gravity is share-to-handle, not publish-to-public. Default visibility is private. The primary action on an artifact is "send to recipient." Public discovery is the opt-in exception, not the default. Growth happens through trusted hops, not algorithmic amplification.

3. Trust-inverted discovery

Most platforms: weak anonymity → users hide → discovery is bad. Stoka: strong anonymity → users expose more about a pseudonym → discovery is rich. The stronger the anonymity primitive, the more a user will opt into discovery — bio, topics, public artifacts, open inbox. This inversion is the entire reason the social layer works.

The Unencumbered Bar (UX quality contract)

If a feature accumulates friction, it gets fixed before it ships. Features accumulate friction by default — the Unencumbered Bar exists to fight it on every change.

Phase roadmap

Stoka platform phases

  1. Stoka v1

    shipped
    useful toVisitors
    2milestones
    • Discovery terminal live
    • Docs, links, products indexed
  2. gatePhase 0 shipped

  3. Solo core

    active
    useful toYou alone, as daily-use tool
    3milestones
    • Artifact capture from pasted conversations
    • Personal library with retrieval
    • Pseudonym(s) for future attribution
  4. gatePhase 1 feels amazing solo

  5. Sharing core

    queued
    useful toInvite round (small, trusted)
    4milestones
    • DM + share-to-handle
    • Rich artifact embeds
    • Inbox
    • Opt-in notifications (in-app, Telegram, web push)
  6. gatePhase 2 stable

  7. Email digests

    queued
    useful toUsers who live in email
    1milestones
    • Verified address digests
  8. gateEnough opt-in public artifacts to seed

  9. Public layer

    queued
    useful toGrowth
    3milestones
    • Opt-in public tier
    • Directory
    • Stoka public discovery
  10. gatePublic layer active

  11. Mod via Stoka

    queued
    useful toAbuse handling
    3milestones
    • Autonomous trigger ruleset
    • Action ladder
    • Transparency page
  12. gateDM stable, demand confirmed

  13. Voice calls

    deferred
    useful toLater
    2milestones
    • Anonymous voice presets
    • Prosody-aware TTS

The critical gate is Phase 1 → Phase 2. Don't invite anyone until you'd use it every day alone.

Specs in this directory

Spec What it covers
stoka-bot.md Stoka the entity — personality, voice, RAG architecture, prompt layers, training pipeline. Now includes "Stoka Across Surfaces" — the one-brain-many-masks pattern.
artifact-platform.md The artifact primitive, kinds, visibility tiers, Bon Appétit granularity ladder, capture flows, solo-first Phase 1
messaging.md Pseudonyms, DM core, share-to-handle, Telegram doorbell, Unencumbered Bar mechanics, Stoka-as-mod, voice (Phase 5+)
stoka-bot-phase2-plan.md Stoka v2 extensions (ambient mode, living system telemetry)
wz-dream-system.md White Zetsu research system
zetsu-shared-brain.md Zetsu fleet shared knowledge layer

Cross-cutting design threads

Pseudonyms as the identity primitive

Every artifact, every DM, every voice call uses pseudonyms — not real accounts. One user → many pseudonyms simultaneously, each with its own bio/topics/discoverability/inbox mode. Pseudonyms are cheap to spawn, cheap to burn. User↔pseudonym link is stored plain (for cross-pseudonym abuse handling), never user-visible. See messaging.md.

Stoka is the connective tissue

One reasoning body, many masks. Same identity/constraints/voice across every surface where the product "talks." Different context layers (retrieved content, pseudonym history, mod triggers) produce surface-appropriate behavior. See stoka-bot.md → Stoka Across Surfaces.

Autonomous moderation with public commitment

Moderation runs as an autonomous Stoka-context. Trigger rules are public. Action ladder is predictable. Every de-anonymization query is logged. Audit log is surfaced publicly. Trust = "we promise" + verifiable trail.

Export is a first-class deliverable

A user's library is theirs. Markdown + JSON export available from Phase 1. The anti-lock-in guarantee is part of the trust story, not a retention problem.

What this platform is NOT

  • Not a feed. No algorithmic timeline. No "for you" page.
  • Not virality-optimized. No follower counts, no like counts, no share counters as primary signal.
  • Not a chatbot. Stoka is the site with a voice, not a customer service agent.
  • Not identity-tied. Real names, real photos, real accounts never appear on the social surface. Pseudonyms all the way down.
  • Not retention-optimized. Per the Stoka Ethos: optimize for usefulness, not engagement. Will recommend OFF-platform if that's what helps.

Living document

This README is updated when the mission evolves. Each phase's spec is its own living document — schema, open questions, and decisions land there.