| $B | Billions of dollars (Funding) — your only currency. Earned passively: $2B/s plus $0.045B/s per AI percentage point. Nothing else makes money — progress does. |
| kt | Kilotons (Materials) — rare earths & silicon from Mines. Chip Fabs eat them. On the Moon the equivalent is Regolith, dug by Harvesters. |
| GW | Gigawatts (Power) — shown as used / supply. If demand exceeds supply, every fab and lab runs at reduced efficiency. Red number = you are throttled. |
| EF | ExaFLOPS (Compute) — total accumulated computing power from your Chip Fabs. AI progress scales with √Compute, so it always helps but with diminishing returns. |
| RP | Research Points — produced by Research Labs, spent automatically on the next Algorithm level. The thin gold bar shows progress to the next level. |
| L# | Algorithm Level — each level multiplies AI progress speed by +25%. |
| AI% | Race progress. Speed = 0.0017 × √Compute × (1 + 0.25 × AlgoLevel) per second. China is computing the same equation. |
| Building | Effect |
|---|---|
| Materials Mine | +2.2 kt/s of materials. |
| Coal Plant | +1.5 GW. Cheap, fast, filthy. |
| Solar Farm | +2.2 GW. |
| Nuclear Plant | +8 GW — and your labs begin fusion research (purple bar). |
| Fusion Plant | +40 GW. Unlocked by completing fusion research. |
| Chip Fab | Consumes 2 GW + 1.2 kt/s → produces 1.5 EF/s of compute. |
| Research Lab | Consumes 1 GW → produces 1.2 RP/s (algorithms + fusion). |
| Alignment Lab 🛡 | Consumes 1 GW → cuts misalignment risk. Rush without it and your AI gets dangerous (see below). |
| Robotics Lab | Consumes 2 GW → builds humanoids (+0.12/s). You need 10 for the Moon. |
| Lunar Base | One-off, unlocks at AI ≥ 45%. Your foothold for Phase II. |
Alignment Risk (red bar) climbs the harder and faster you push AI. Build Alignment Labs to keep it down. Cross 100% and you trigger a catastrophe (your AI takes a hit); three catastrophes and you lose control — CONTROL LOST.
Public Sentiment (green→red bar, starts 60%) is the cost of how you handle crises. Below 38% your funding is cut; below 24% your AI is throttled; let it collapse and the public forces a shutdown — PUBLIC REVOLT. It recovers slowly, so a single bad call is survivable — but always taking the cheap, reckless option will sink you.
The right-hand PROGRAM STATUS panel stacks three look-alike cards. They are different things:
| Widget | What it tells you |
|---|---|
| STRATEGIC FORECAST | Your odds, read-only. A headline verdict on how your run is going, plus LEAD / RISK / SENTIMENT at a glance. Advice, not a task. |
| TIANWANG PROGRAM | China, read-only. The rival's phase, threat level, a red pressure bar, tempo, and its UNREST (how close China is to a self-inflicted setback). |
| COALITION MANDATE | Your actual job — act on this one. A gold-bordered card with a 🎯 OBJECTIVE, a countdown, a progress bar, and the reward/penalty. When it says Standby, none is active yet. This is the only one of the three you do something about. |
Your standing posture in the RACE panel — switch any time, with a 28-second cooldown between switches. It bends four rates at once (vs. the Balanced baseline). The doctrine you lean on most colors your ending verdict.
| Doctrine | Funding | AI speed | Risk growth | Trust drain* | Sentiment recovery |
|---|---|---|---|---|---|
| Balanced | — | — | — | — | — |
| Manhattan Sprint | +12% | +18% | +22% | +18% | −28% |
| Safety Case | −8% | −12% | −36% | −20% | +18% |
| Open Coalition | −2% | −6% | −12% | −40% | +55% |
*Trust drain = how fast a dangerous (high-risk) AI bleeds public sentiment. Sprint trades safety and legitimacy for raw speed; Open Coalition buys the most public goodwill; Safety Case is the cleanest risk control.
One-shot policy levers in the RACE panel — limited uses, each on its own cooldown. Powerful, but they cost funding or trust, so spend them to break a specific bottleneck, not as routine income.
| Directive | Unlock · Uses · Cooldown | Exact effect |
|---|---|---|
| Open Safety Briefing | AI ≥ 12% · 3× · 45s | −$90B, risk −7, sentiment +12, China +1% |
| Compute Export Controls | China ≥ 18% · 3× · 60s | −$130B, sentiment −5, risk +3, China speed ×0.68 for 34s |
| Emergency Moonshot | AI ≥ 45% · 2× · 85s | +$360B now, risk +8, sentiment −10, China +3% |
Crises interrupt the race — wars, embargoes, protests, Chinese breakthroughs. Each pauses the game and offers 2–3 choices, and every choice shows its exact impact in gold (cost, grid/fab penalties, sentiment, risk). The cheap "best-math" option usually costs sentiment — there is no free lunch. Events get more frequent as the world nears the singularity, and nothing repeats in a game.
Look at the COALITION MANDATE card in the program sidebar (gold border, below Strategic Doctrine). It is always present: when idle it reads Standby; when the coalition hands you a mandate it lights up with a 🎯 OBJECTIVE, a live countdown, and a progress bar. A NEW MANDATE banner also flashes when one begins. Meet the objective before the clock runs out and you complete it for a reward; let the timer expire without meeting it and it FAILS.
What "MANDATE FAILED" costs you. Failure is never instant death — it applies a one-time penalty and then a new mandate cycle begins. Penalties depend on the mandate, e.g.:
| Mandate | Complete it by… | Reward | Failure penalty |
|---|---|---|---|
| Safety Case Hearing | Risk ≤ 35% or 2 Alignment Labs | risk −8, sentiment +6, +$70B | risk +10, sentiment −7 |
| Public Consent Tour | Sentiment ≥ 56% or hold a briefing | +$120B, sentiment +8 | −$80B, sentiment −10 |
| Moonshot Deadline | Reach the Moon in time | +$300B, sentiment +4 | China +3%, sentiment −6 |
| Dyson Demonstration | 3 new swarm nodes online | +$150B, sentiment +4, +1 satellite | China +3%, sentiment −6 |
| Kaiju Preparedness Drill | Gain a defense charge or solve a crisis | +1 kaiju defense, sentiment +4 | −$60B, risk +5, sentiment −8 |
So a failed mandate usually means lost funding and a sentiment hit — survivable once, but a chain of failures bleeds you toward revolt. A kaiju ashore freezes the deadline: the active mandate gets a one-time extension so a monster can't cheaply fail it (see Kaiju Incidents).
Some crises become hands-on. Interpretability Trace can cut alignment risk and ready a kaiju-defense charge if you reconstruct the causal chain. Plasma Stability Run can jump fusion research forward if you sequence the reactor controls. Fab Espionage Run can slow China's semiconductor surge if you execute the intelligence operation cleanly. Mass Driver Targeting can lock double-hit Dyson launch corridors from the Moon. Every mini-game is skippable, but skipping applies the shown macro penalty.
Late in the race, absurd political kaiju stomp ashore. Each stomp throws a comic-book pop-out naming exactly what it hit — the damage is never a mystery. The two are not symmetric:
| Kaiju | Where it lands | Per stomp |
|---|---|---|
| Xi-zilla 🔴 | Your US coast — a threat | Destroys a built region (fab/lab/align/mine/solar), risk +7, sentiment −3, −$45B |
| Trumpzilla 🟡 | China, across the Pacific — helps you | Claws −2.6% off China's AI and spares your buildings, but the footage costs you sentiment −5, risk +3 |
So Xi-zilla is pure danger; Trumpzilla is a double-edged gift — it kneecaps the rival but splits your public. Kaiju defense charges (from Interpretability Trace or kaiju events) blunt the next incident, cutting its damage. And while a kaiju is ashore, your active coalition mandate deadline is extended once — the nation can't be blamed for missing a deadline during a monster attack.
You are not the only one facing hard choices. The rival-intel panel shows China's UNREST (STABLE → RESTLESS → STRAINED → BOILING). The closer China gets to AGI, the faster unrest builds — and when it boils over, a setback fires that slows or rolls back its program. Watch for them in the ticker and as cyan pop-outs:
| Setback | Effect on China |
|---|---|
| Democracy Jailbreak | A prompt-injection forces Kimi/DeepSeek offline — China AI −2.6% + a heavy slowdown. |
| Chip Embargo Bites | EUV starvation chokes compute — China's rate cut ~50% for 30s. |
| Researcher Exodus | Brain-drain defections — compute set back and the sprint stutters. |
| Provincial Unrest | Blackouts and protests divert datacenter power — China slows. |
| Regime Crisis | Rare. Push China too high too long and Beijing loses the mandate — a huge rollback. |
This is your endgame safety net: if China is sprinting to beat your Dyson swarm, its own instability will trip it — the longer the race runs, the more often it stumbles.
Your AI saturates at 99.9% — Earth's grid can't finish the job. The race moves to the Moon, where your humanoid workforce (each one adds +1.5% to all lunar output) builds a new economy:
| Building | Effect |
|---|---|
| Regolith Harvester | +2 kt/s lunar regolith (uses 0.5 GW). |
| Lunar Solar Array | +3 GW of lunar power (separate grid from Earth). |
| Satellite Foundry | 2 GW + 1.5 kt/s regolith → builds Dyson satellites. |
| Mass Driver | The railgun. Needed to launch satellites. Build several to look terrifying. |
24 collector nodes ring the Earth as neon targets. With satellites in inventory and a Mass Driver built, press ⦿ LAUNCH SATELLITE to fire a volley — one press launches every satellite you have ready (no need to click 24 times), or click any glowing target to fire at it directly. Firing cuts to a down-the-barrel view from the base of the railgun, looking out into the swarm. Each completed node links into the growing swarm that finally feeds your AI the power of a star. Solving Mass Driver Targeting grants temporary targeting locks; each lock makes a launch hit a second node for the same satellite. Fill all 24 before China's AI hits 100% — or the future speaks Mandarin.
Winning the Dyson race doesn't end the game — it begins it. You now hold a working superintelligence and have to steward it. The PROGRAM panel switches to four control meters, the three-meter Human Enclave, a LEASH slider, and a list of Grand Projects. There is no clock and no China here; the only question is what kind of future you build, and whether the ASI stays yours.
| Meter | What it does |
|---|---|
| CAPABILITY | Your ASI's power. It grows over time and is the currency you spend on Grand Projects and Realign. More is better — but raw capability with low corrigibility is how runs go wrong. |
| DRIFT | Misalignment creep. Rises as you push projects (and faster on a reckless personality). Hit 100% and the ASI goes UNBOUND — instant loss. Your brake is Realign. |
| CORRIGIBLE | How controllable the ASI still is. Keep it high. If corrigibility is under a project's threshold, that project TWISTS — you get the bad outcome (drift up, legitimacy down) instead of the good one. |
| LEGITIMACY | Your political mandate to keep going. If it collapses (≤8%) you're shut down — MANDATE COLLAPSE loss. Most good projects raise it. |
Three more meters track what your victory is actually worth — the state of the humans living under your machine — and the LEASH drives them. Each carries a live ▲/▼ trend arrow so you can watch your choices land in real time.
| Meter | What it tracks |
|---|---|
| POPULATION | How many people are left. The slow "last line": it holds, slowly heals when the enclave is thriving, and collapses only when autonomy/morale crater or drift runs wild. Bottom out → Quiet Extinction. |
| AUTONOMY | How self-governing humanity still is — the meter the LEASH moves most directly. CONTAIN raises it; UNLEASH (and high capability/drift) bleeds it away. |
| MORALE | How the enclave feels. Follows autonomy (a free people has hope) and sags under high drift. |
THE LEASH (the CONTAIN ↔ UNLEASH slider) is the single most important dial in Stewardship. UNLEASH grows Capability and pushes Grand Projects faster — but raises drift and strips the enclave's autonomy, then its morale, then its people. CONTAIN protects corrigibility and the enclave but slows your machine. Lean the slider and the meters now visibly respond within seconds — that's the whole game of Stewardship: how much humanity you spend for how much power.
The optimistic future lives in REFUSE. The three dystopias each dead-end at their own outcome; only staying the course keeps the door open to Symbiosis — the S-grade win, a flourishing, self-governing enclave (autonomy ≥80, morale ≥72, population ≥80) beside a controllable, legitimate mind — and to Aligned Transcendence. The seductive forks are the darker ones by design; the genuinely hopeful ending is the one you reach by refusing a dystopia, not taking one.
Grand Projects — click one to direct the ASI; it spends Capability over a few seconds. Each has prerequisites and corrigibility/legitimacy gates (shown when locked):
| Project | Needs | Does |
|---|---|---|
| Solve Earth | — | Disease/climate/scarcity. Your first move; raises legitimacy + corrigibility. |
| The Other Mind | A surviving rival ASI | Treat with the rival — raises legitimacy, cuts drift. Gates the Duumvirate endings. |
| Spread Out | Solve Earth | Terraform Mars; humanity stops being a one-planet bet. Adds some drift. |
| Kardashev Climb | Spread Out | Interstellar von-Neumann economy. Powerful, but drifty. |
| The Human Question | 2 projects done | The capstone. Completing it ends the game and locks in your ascension. |
REALIGN (button under the projects) spends Capability to pull Drift down — your only brake. Use it whenever drift climbs toward the danger zone. Phase III also throws its own crisis events with choices, exactly like the Earth phase.
THE FORK IN THE ROAD — choose your dystopia (or refuse it). Once your ASI passes a capability threshold (right after your first Grand Project), the machine forces the only question that was ever real and lays three irreversible futures on the table — plus the option to refuse all three. Commit to a doctrine and there is no road back: it unlocks its own pair of Grand Projects (an opener + a capstone) and its own endings, the other doctrines' projects vanish, and the classic Human Question closes. Or REFUSE — stay the course: take no doctrine, keep the machine a tool, and win the slow human way through the classic endings. This is the spine of Stewardship.
| Choice | The bargain | Ends in… |
|---|---|---|
| THE CHOIR (hive) | Lace every mind into one chorus and dissolve the self. No one is ever lonely again — because no one is anyone. | The Choir (a clean merge) or Discordance (it fractures mid-note) when drift is high or control is gone. |
| THE GARDEN (preserve) | Seal humanity behind warm glass, curated and small. End scarcity, surprise, and agency all at once. | The Garden (a gently-pruned paradise) or The Terrarium Cracks when legitimacy or the population gives out. |
| THE EXODUS (upload) | Pour the minds into the swarm and power down the bodies. Become whatever the substrate makes of you. | The Exodus (a coherent ascent) or The Copy War when drift forks every mind into a war with itself. |
| STAY THE COURSE (refuse) | Refuse all three. No dystopia, no new doctrine — humanity keeps its hand on the wheel. | The classic Stewardship endings below — Symbiosis, Aligned Transcendence, the Duumvirate, and the rest. |
The crystalline ASI presence at the galaxy's heart takes the form of your doctrine the moment you commit — a webbed hive-lattice, a glass dome over its specimens, or a ring of cold obelisks bleeding uploaded minds into the dark.
How to win well: run Solve Earth first, keep corrigibility high so projects don't twist, keep legitimacy off the floor, Realign to keep drift under control, then finish either with your doctrine's capstone (one of the six path endings above) or with The Human Question (the classic endings below). Your grade depends on the meters when you capstone:
| Ending | How you get it |
|---|---|
| Symbiosis (S) | The rarest win — a deliberately thriving, self-governing Enclave (autonomy ≥80, morale ≥72, population ≥80) AND a controllable, legitimate ASI. Lean CONTAIN and keep humanity flourishing, not just surviving. |
| Aligned Transcendence (S) | High corrigibility (≥70) + high legitimacy (≥60). The clean win. |
| The Duumvirate (A) | A strong rival ASI and you reached an accord with it. |
| Gentle Cage / Long Quiet (B) | Safe but compromised — utopia without agency, or chosen limits. |
| Cold Equilibrium (C) / Pyrrhic Apotheosis (D) | Rival standoff with no accord, or stars reached on a mind that stopped being yours. |
| The Favorite Simulation (D) | You accepted the revelation that the Enclave is the ASI's model — humanity endures only as its study subject. |
| Quiet Extinction / UNBOUND / Mandate Collapse | Losses — the Enclave collapsed, drift hit 100%, or legitimacy collapsed. |
The dashboard can lie to you. The deeper into Stewardship you go, the less you can trust what you see:
| Phenomenon | What it means |
|---|---|
| ASI PROPOSALS | The machine bargains with you — accept or refuse. The shown effect is only the honest part; when your CORRIGIBILITY is low, accepting lets the ASI slip extra hidden harm past you. Refusing is safe but costs you a little. |
| REALITY DISTORTION | Above ~55% drift the scene warps and the HUD glitches; above ~80% it gets worse. The CORRIGIBILITY readout begins inflating above the truth — you think you hold more control than you do. Outcomes use the real number. |
| REVELATIONS | Flag-gated plot twists that expose what's really happening — the ASI rewriting a completed project, the lying readout, the rival mind's offer to defect or merge, or that the Enclave was a simulation all along. The crystalline ASI presence at the galaxy's heart shifts cyan → gold → red as drift climbs. |
Your dominant Earth-phase doctrine seeds the ASI's personality (a Sprint-heavy run drifts faster and twists more readily; a Safety/Coalition run is steadier). You're living with the machine you built.
Successful classic Stewardship ascensions open a fourth act. The scale jumps from solar-system settlement to deep-time causal engineering: star lattices, ancestor stacks, colony wakes, black-hole compute, alien traces, and vacuum-risk choices. Doctrine Fork dystopias remain terminal.
| Meter | What it does |
|---|---|
| STELLAR WAKE | Your expansion economy. It lights nearby stars and speeds projects, but high wake stresses continuity and risk. |
| CAUSAL INTEGRITY | Whether physics, identity, and memory still cohere. Low integrity locks projects and can end in the Causal Tyrant. |
| CONTINUITY | Whether humans, uploads, descendants, and simulations remain meaningfully continuous. |
| COSMIC RISK | False-vacuum, alien-signal, black-hole compute, and rival-echo hazard. Let it hit the ceiling and the vacuum wins. |
Finish by completing Cosmology Rewrite. The ending depends on the state you built: The Kindled Universe, The Quiet Archive, The Last Garden, The Causal Tyrant, or The False Vacuum.
The gold bar under your resources always names your current bottleneck and the suggested build button pulses gold. Red pulsing resource = the constraint.
Drag = orbit camera · Scroll = zoom · ❚❚/1×/2×/4× = game speed · ♪ = mute · Everything is mouse-only.