A Risk-like conquest game played one tick per day, for a private group of friends.
Nobody plays it live. Each day you submit standing orders through a web app, and at midnight everyone's orders resolve simultaneously. Combat is fully deterministic — there are no dice.
Reinforcements come from two places beyond territory income:
- Real-world actions. Post a workout photo in Slack; two friends react; you get soldiers. This is what makes the game an accountability device.
- Prediction market wagers. Stake soldiers on real Kalshi markets closing that day. This is what replaces dice — variance you can actually reason about.
A sample of what a day's slate looks like in practice: Bitcoin's close, the price of gold, a film's Rotten Tomatoes score, a Uruguayan football match, and a WTI crude settlement.
| Document | What it is |
|---|---|
docs/superpowers/specs/2026-08-09-riskety-rekt-design.md |
The spec. Every rule, and why it is that rule. |
HANDOFF.md |
Current state, architecture, and the rules a newcomer gets wrong |
codemaps/ |
Generated architecture maps — module graph, engine pipeline, schema, jobs, web. Regenerate with /update-codemaps; don't hand-edit. |
docs/superpowers/reviews/2026-08-11-balance-run-modules.md |
What 10,000 simulated seasons say about the economy |
docs/superpowers/plans/ |
Implementation plans. Each carries a "Spec deltas" section recording where reality corrected the design. |
The spec is the interesting document. Most of its rules exist because a reviewer or a simulation found the obvious alternative was exploitable — a both-sides hedge worth a risk-free +10% a day, a 1-troop feint that voids a 100-troop assault, a slate of five bets that are secretly the same bet.
The pure rules engine (module-dispatched: markets, IRL and the veto are pluggable
mechanics, and a voted daily rule catalogue rides the same hooks), the offline
season simulator, the Kalshi market adapter, the Slack ingress, the midnight tick
runner, the order-entry CLI and the player web app are built and tested.
1,034 tests, none of which touch the network — test/no-network.ts replaces
fetch in every test run, so that is enforced rather than asserted.
The player app is built: sign in with /login in Slack, act on a real map,
orders autosave and lock at midnight. Each morning the bot posts a rule vote; the
winner applies to that night's tick. What still blocks a competitive season is
snowballing — see "Not built" in CLAUDE.md.
npm install
npm test # 1,034 tests
npm run typecheck
npm run sim # 2,000-season balance run, ~2s
npm run sim -- Slacker Blitz GymRat # custom policy roster
# slack — see deploy/README.md
npm run roster:add -- U01ABCDEF f1 "Ada L."
npm run roster:list
npm run slack # the events bot, a long-running servicesrc/engine/ pure — zero dependencies, no I/O, no clock, no randomness
src/sim/ drives thousands of synthetic seasons through the engine
src/adapters/ the only code that speaks HTTP; parsing split from networking
src/slate/ pure slate selection
src/store/ SQLite (node:sqlite, WAL) — slates, settlements, Slack ingest
src/slack/ Bolt ingress, approval derivation, and Block Kit rendering
src/jobs/ the 08:00 publish, the 30-minute poll, the recap, and their CLI
The engine is one function — resolve(state, orders, context) → GameState — and it
imports nothing outside its own folder. A test enforces that, along with a ban on
Date.now(), Math.random() and new Date(). Time and randomness enter as arguments.
That purity is why the simulator can exist, and the simulator is the only reason anyone knows the economy isn't broken.
The tick never touches the network. Slack approvals arrive continuously by webhook and market settlements are written by a poller, both landing in SQLite well before midnight. A Kalshi outage at 23:59 cannot stall the season.
See deploy/README.md for running the market jobs and the Slack bot.
MIT © 2026 Sean Perkins