SWII CoPilot
A governed AI workspace for federal capital formation. Upload one file to any AI — it becomes a copilot that computes what is legal, feasible, and what it costs. It is governed never to recommend, never to pick, and never to choose for you.
Two paths to the same destination. Start with the teaching game to learn the method on a safe board. Or go straight to Live Play with your real evidence. Same engine, same rails, same governance — the game teaches the copilot you'll use on your own data.
The product ladder
Everything on this platform is downloadable content. Two architecture paths, one governed method.
Individual prompt modules. Plug into any active AI project — capsule or copilot.
Self-expanding project capsules. Fine-grained skill control — add one module at a time.
Teaching game. Synthetic board, practice system. Learn the copilot before using it on real data.
All-inclusive copilot. Your real evidence, your real organization, governed computation in one engine.
Two paths, same destination
Fine-grained path (V1 + V2): Download an AI Starter for your vertical. Add Skills from the library one at a time as you need new capabilities. You control what's installed and when.
All-inclusive path (V3 → V4): Learn the method with the teaching game. Graduate to Live Play — the full copilot engine in one file. Skills from V1 can still be carried in as reference alongside either path.
Starter Mode v3 — Teaching Game
An AI-run tabletop strategy game about positioning before the deadline. Fictional town, synthetic board, dice, sealed scenarios — a practice system where the copilot teaches you the method by running the table.
Every rule you learn here is a rule the Live Play copilot enforces on your real data. The game is the training ground.
Who it's for
Ages 14 and up for independent play · Ages 10 and up with a Table Operator (an adult or experienced player on the machine side). One to four players per table.
What's in the capsule
- →The AI Game Master — governing instructions your AI reads first. It computes what's legal, what it costs, and what it risks — governed never to choose for you.
- →A sealed scenario — "First Quarter in Harborvale," wholly fictional, four rounds.
- →The spoken teach — read aloud at the table; first turn in minutes.
- →The table-read sheet — what off-turn players do.
- →A Setup Receipt — dated proof the install worked.
You'll also want one ten-sided die. If your AI can run code, it can roll for the table.
How install works
Download the capsule.
Add it to a fresh AI conversation or project.
Say START GAME. The Game Master speaks its custody promise.
Live Play v4 — All-Inclusive Copilot
The full governed copilot in one file. Real organization, real evidence, real government records — carried by you, computed by the engine. Platform-aware setup, evidence-classified org profile, field-specific questions for your vertical, and a Shopping List that tells you exactly what to bring and what it blocks.
The governed copilot turns out to have been what the game was teaching all along.
Who it's for
Adults managing real federal capital formation work — grants, contracts, policy, finance, sponsored research. One organization per install.
What's in the copilot
- ✓Platform detection — the copilot identifies what your AI can do, adapts every instruction to your environment
- ✓Organization interview — structured profile, destination, constraints, commitment authority
- ✓Evidence classification — every org fact marked DOCUMENTED, ATTESTED, or UNKNOWN
- ✓Field profile — conditional questions for your vertical (govcon, grants, policy, finance, cross-program)
- ✓Shopping List — every gap named, what it blocks, where to resolve it, how to carry the answer back
- ✓The courier interface — carry snapshots, records, and org changes with full provenance
- ✓Update cycle — HEAR → ASK → CHOOSE → CHECKPOINT on your real board
- ✓Provenance on everything — source id, record date, retrieval date on every receipt
- ✓Freshness tracking — board age on every receipt, STALE flag when evidence is overdue
- ✓Seven failure classes — named incident protocol from unsourced claim to fabrication
- ✓Checkpoint and recovery — print-and-restore from receipts, never from memory
- ✓Screenshot support — paste a photo as a question, as evidence, or as an error report
- ✓The boundary — computes everything, recommends nothing, commits nothing
What carries from Starter Mode
- ✓The governance model (computes, never recommends)
- ✓Receipts as state — the transcript is the save file
- ✓Runway arithmetic and the three measures: time, people, money
- ✓The organization book (now with provenance fields)
- ✓The two-line scoreboard: organization status + decision evaluation
What changes from Starter Mode
- ×Dice die. Reality resolves gates — a notice publishes or doesn't, an agency awards or doesn't.
- ×Sealed scenarios die. Replaced by the courier interface — you carry real snapshots and government records.
- ×Fictional orgs die. Your real organization: actual people, budget, constraints, goals — each fact evidence-classified.
- +Platform detection. The copilot identifies your AI environment and adapts.
- +Evidence classification. Every organizational fact marked by evidence quality.
- +Shopping List. Gaps named with what they block — your acquisition plan.
How install works
Download the capsule.
Add it to a fresh AI conversation or project.
Say START SESSION. The copilot detects your platform, runs the setup interview, and emits your Shopping List.
Can I use Skills from the library alongside Live Play?
Yes. Skills from V1 can be carried into a Live Play project as reference material. The copilot remains the single governing engine — Skills inform your thinking, not the copilot's instructions. This is the difference between the two paths: AI Starters (V2) let Skills govern; Live Play (V4) lets them advise.
The human-AI dyad is always free.
You carry evidence manually — screenshots, downloads, copy-paste from government sources. The copilot computes against what you bring. This is the free tier, and it is the full product.
Automated data synchronization for freshness — where the system retrieves evidence for you — is a future paid tier. The copilot's computation is identical either way.
The honest print
Playtest release. Both tiers are under active development; rules carry their draft status inside the capsule. Starter Mode band edges are provisional pending playtest. Live Play is proposed and not yet ratified.
Starter Mode is a teaching game played on a synthetic board — every organization, program, and record in it is fictional. Live Play operates on real evidence you supply — but neither tier offers advice: not legal, not financial, not grant, contracting, or investment advice.
Compatibility: built for file-capable AI tools including ChatGPT, Claude, and Gemini; features and results may differ by provider, model, plan, and file limits.
Self-custody, free. Both capsules are yours: no account, no telemetry, your state lives in your own AI project. One organization per Live Play install.
Doing nothing is a move, and it's priced. So is winning more than your organization can carry. Don't blow up.