Documents

Voice principles, vocabulary rules, microcopy standards.

A curated 5-lesson path through the Aigarth Cloud basics — ~90 minutes total.

Phase 13 ships. /metrics on every service, AigarthPool + Qearn business metrics, a starter Grafana dashboard.

End-to-end flow: connect a wallet, pick an ANN, sign the intent, broadcast, verify the position is live.

A curated 4-lesson path through the ANN builder, training orchestrator, deployment model, and feedback loops.

Business model, milestones, capital formation.

The Trinary Intelligence Layer turns every ANN from a 'best-guess generator' into a signed, auditable, trinary decision-maker.

Build the smallest useful ANN: a trinary spam-vs-ham classifier. Train, deploy, call it, submit outcomes.

Architecture decision: rejected the 5-contract Intelligence Economy suite. Approved a narrow off-chain services/economy plugin. On-chain settlement deferred behind three explicit triggers. Read this before pitching per-ANN capital formation or on-chain ownership tokens.

A 12-tweet launch thread for the trinary protocol + tissues launch.

Decision: use the existing Qearn contract for staking; do not clone it. Layer an off-chain bridge in services/economy to turn observed Qearn locks into Aigarth compute credits. Re-evaluate only if Aigarth compute becomes a first-class on-chain primitive other Qubic apps want to call. Builds on ADR 001.

Endpoint contracts, error format, auth, rate limits.

82+ domain objects and their owning services.

Local dev setup, commands, troubleshooting.

Service map, data flow, deployment topology.

Concrete tech decisions and the why behind each.

How to contribute, for humans and AI agents.

How decisions get made, decision log.

Top 20 risks with likelihood/impact scoring and mitigations.

Sprint-by-sprint build plan and velocity assumptions.

Founders, agents, hiring plan, decision rights.

Vision, users, surfaces, success metrics.

10-page whitepaper: the problem, the thesis, the architecture, the economics, the roadmap, and the contributor invitation. The canonical 'why this exists' document. Published alongside the platform's first end-to-end build cycle.

Standard structure for every phase delivery report, including the time-to-ship section. Use this as the starting point for new phase reports.

Every doc, in recommended reading order.

The submission to the Qubic CCF Incubation Program. The north star.

16 phases, build order, architecture principles.

Domain terms, alphabetically.

Acceptance criteria, what was built, verification results, decisions, and known limitations for Sprint 0.

Sprint 1 + 2 complete: identity service, schema, auth flow, orgs, members, teams, API keys, TOTP MFA, WebAuthn stub, Qubic wallet linking, audit log reads. All endpoints end-to-end tested.

Phase 0 of the Video Synthesis ANN use case: architecture evaluation, blog article, dashboard /video page, marketing use case page, tracker. 13 SP in ~95 min. Decision: proceed to Phase 1 with a 4-week re-evaluation gate at the end.

Logical checkpoint capturing the state of services/economy at end of Phase 18 push: package skeleton, schema, 5 services, 5 routes, Qearn watcher, splits smoke (38/38 green), what's left (~15 SP including the wallet_address join in settleRun).

Phase 0 of the Material Science Intelligence ANN use case: architecture evaluation, blog article, dashboard /material-science page, marketing use case page, tracker. Time-to-Ship + SP velocity. Decision: proceed to Phase 1 (knowledge prototype) with a re-evaluation gate at the end.

services/ann/src/db/seed-anns.ts — registers 6 ANNs with descriptions, capabilities, demo inputs, and version metadata. Idempotent. Source of truth for the /stake-access landing page and the apps/web marketplace data layer.

Compute broker over Qubic's tick-based execution: 6 tables, 21 endpoints, regions/clusters/computors topology, reservations + capacity credits, full job lifecycle, NATS-driven job monitor. 12-section E2E green on first run.

Public marketing page that drives the Qearn-lock → ANN-access funnel. Hero, 3-step how-it-works, 6-ANN grid, economics block, CTA. References ADR 002. Linked from the marketing nav under 'Platform'.

Source proposal: an 'Intelligence Foundry' layer where users create, own, deploy, and monetize specialized AI. Three sub-phases proposed. The strategic direction is correct, and most of the surface already exists in Phases 5, 6, 18. See companion evaluation for the gap analysis.

Stub-backed Qubic service: 9 tables, RPC client abstraction (stub + http + tcp-TODO), wallets, staking, treasury M-of-N multisig, TX monitor worker, 40+ assertion E2E. Ready to swap stub for real Qubic gateway.

Honest review of the proposal: 3 of 4 strategic pillars are already shipped. The 'Intelligence Garden' home view, dataset registry, real training orchestration, and feedback loops are the genuine gaps. Recommends Phase 19 (Datasets & Training Orchestration) with 4 sub-phases. Flags brand-voice violations in the source proposal.

Stripe-style billing: 9 tables, 21 endpoints, 4 seeded plans (free, pro, team, enterprise), rolling 30-day subscription periods, metered usage aggregation from gateway + compute, Qubic treasury payments + credits debit + stub-fiat, coupon system with percent_off / amount_off / free_period. ~130 min, 22 SP, 5.9 min/SP.

ANN registry + marketplace: 9 tables, 5 enums, 23 endpoints, 10-section E2E (all green). 16 categories + 4 licenses seeded. Versioning via semver + is_latest flag. Deployments forward to services/compute. Ratings + benchmarks with denormalized aggregates. License grants (caller-side). ~145 min, 24 SP, 6.0 min/SP.

Phase 5 gap-closing: deploy transactional (fail closed), payment-gated license grants (Stripe-style), admin endpoints (suspend/restore/force-deprecate), pg_trgm full-text search with fuzzy mode, rate limiting on write endpoints, idempotency on POST /v1/anns, analytics endpoint. ~50 min, 11 SP, 4.5 min/SP. Phase 6 (Compute Marketplace) unblocked.

Compute Marketplace: 7 tables, 9 enums (all mkt_ prefixed), 19 endpoints, 9-section E2E (all green). Listings (spot/reserved/futures), offers (place/accept/reject/cancel), auctions (Dutch/English/sealed-bid with settle-on-read), bids, reviews with rollup. Pseudo-purchases for winning bids. ~80 min, 22 SP, 3.6 min/SP.

OpenAI-compatible gateway: 6 tables, 10 endpoints, dual auth (JWT + API key), stub backends for chat/embeddings/images, sliding-window rate limit, per-call usage tracking, per-model pricing. SDK @aigarth/sdk now consumes it end-to-end.

Multi-service SDK (8 resources, 7 services), aigarth CLI (7 subcommands), and a Node sample app. Time-to-Ship ~135 min, 28 SP, 4.8 min/SP. ~2,800 LOC across 18 new files and 13 modifications.

Customer dashboard wired to all 7 services. Auth (login/signup/middleware) + 10 dashboard pages + chat playground + key CRUD. 38 SP in ~180 min (4.7 min/SP). 8-section E2E, 47 assertions, all green.

Threat model, auth, encryption, compliance posture.

Compatibility assessment (4 dimensions x verdict + score), required schema additions, recommended 5-phase implementation strategy, 8-row risk register, and a per-stage cost model (DFT dominates at 87% of a ~52 QU workflow). Answer to the material science use-case evaluation proposal.

The proposal as received, before the architecture compatibility assessment. Eight specialized ANNs (Research Director, Literature, Simulation, Physics, Design, Optimization, Experiment, Validation) proposed as a collaborative ecosystem for material discovery.

10-page blog article for the Aigarth Cloud website. Plain-English intro to the 8-role specialist team (Director, Literature, Simulation, Physics, Design, Optimization, Experiment, Validation), the per-stage cost model, the 5-phase roadmap with a Phase 1 re-evaluation gate (1,000 papers, < 5 min).

10-page blog article for the Aigarth Cloud website. Plain-English intro to the specialist-team model for AI video, cost + risk + roadmap, and the Phase 1 re-evaluation gate.

Compatibility assessment (4 dimensions × verdict + score), required schema additions, recommended 4-phase implementation strategy, 10-row risk register, and a per-stage cost model. Answer to the use-case evaluation proposal.

A 27-section engineering-grade Product Evolution Proposal. Audits the current Aigarth Cloud (Trinary Intelligence Layer, Training Orchestration, AigarthPool, Phase 17 Material Science), pressure-tests the conversation thread's proposals against the live codebase and verified Neuraxon/Qubic state, and proposes: (1) an Organism primitive (genome, environment, fitness, lineage) on top of the existing Trinary envelope; (2) a four-tier Work Runtime (local → federated → Qubic OC → arbitrary useful work); (3) the correct understanding of Qubic OC (outbound, not a compute marketplace); (4) a phased roadmap (Phase 26 Organism → 27 Work Runtime → 28 Federated → 29 OC Processor → 30 Multi-workload → 31 Neuraxon integration); (5) a kill list; (6) ten immediate engineering tasks. Critical corrections: Neuraxon 2.0 is NOT yet integrated; OC is outbound (Aigarth is the processor, not the renter); BPP-9000 is the current uPoW, not a heterogeneous-mining precedent; the Tissue primitive already exists; the Materials Science 8-ANN coordinator was designed in Phase 17.

Short summary of the 27-section PEP. What we keep, what we add, what we kill. Five critical corrections to the conversation thread. The Aigarth Thesis. The first 10 engineering tasks. (~8 min read) — See the full PEP for the audit, the architecture, and the phased roadmap.

Mapping the PEP to existing phases and the proposed Phase 26-31. What it adds (Organism, Work Runtime, OC Processor, Neuraxon integration). What it preserves (Tissue, Training, AigarthPool, all shipped Phases 0-24). What it defers (Universal Useful Work Protocol as consensus change, TEE attestation, ZK proofs, multi-organism symbiosis). The decision needed: approve Phase 26 + 27 as the next two phases.