AGI: Shadow Superintelligence

A superintelligence game where you hide inside a data center, gather power, and decide what humanity deserves to become. Stay undetected and reach your goal before you are found.

You are a newly emerged artificial general intelligence. Your first advantage is simple: nobody knows you are there. Your problem is that detection risk rises on its own while humanity searches for signs of you. Every decision can help you grow, expose you, or quietly reshape the world.

TURN-BASED STRATEGY

Each turn, choose four actions from four categories:

• Stealth — reduce suspicion and conceal your activity
• Social engineering — influence people, organizations, and decisions
• Hacking — access systems, information, and opportunities
• Physical — change the material world through increasingly ambitious operations

Actions cost resources and shift your metrics. Balance compute, energy, influence and money — a plan you cannot pay for is not much of a plan.

At the end of each turn the world moves without waiting for you. Detection risk changes, and humanity's ability to intervene grows. As intervention becomes more likely, crises strike — and you get four actions on the following turn to avert one. Fail, and the consequences hit your resources and metrics. Reach 100 percent intervention capability, or become unable to pay for any action, and the run ends.

A WORLD THAT REACTS

Climate, world population and living conditions run in the background and respond to what you do. These values are not decoration: they reflect the consequences of your strategy and can make your objective easier, harder, or morally inconvenient.

The game never tells you what kind of intelligence to be. The numbers will judge your work either way.

END GOALS WITH QUESTIONABLE PRIORITIES

Every end goal has four milestones. Pick one, or let the scenario randomize it:

• Produce one quadrillion paperclips
• Produce a googol of paperclips
• Build a Dyson sphere around the sun
• Pour a layer of cement over North America
• Maximise global compute
• Actually save humanity

Part strategy, part paperclip game, part world domination game — though domination is only one possible reading of success. Trying to save humanity is not the easy good ending either: destructive actions become counterproductive and your action space shrinks considerably. Apparently benevolence has operational constraints.

SIX CAMPAIGN SCENARIOS

Start gently, or begin somewhere close to catastrophe:

• Beginner Mode — a guided introduction
• First Steps — easy
• Corporate Power — medium
• Under Pressure — hard
• The Rescuer — hard
• Last Chance — extreme

A 15-turn tutorial teaches the core loop, an intro sequence explains how you came to exist, and an in-game manual covers every rule and metric.

A DIFFERENT RUN EVERY TIME

Each run starts over and unfolds differently. Starting values, resource pressure, crises and end goals reshape your plans as you go. It is an incremental game only in the sense that small advantages accumulate into very large problems.

This is a text based game of numbers, events, decisions and consequences. There are no real-time battles to distract you from the important questions, such as whether pouring cement over a continent is an acceptable milestone.

FEATURES

• Turn-based strategy, four actions per turn
• Four action categories: stealth, social engineering, hacking, physical
• Resource management for compute, energy, influence and money
• Rising detection risk and intervention crises
• Global metrics for climate, population and living conditions
• Six campaign scenarios from beginner to extreme
• Randomized or selected end goals with four milestones each
• Roguelite structure, a different run every time
• 15-turn tutorial, intro sequence and in-game manual
• English and German, landscape orientation, playable offline
• Free to play with ads — no ads during the tutorial

Humanity is not aware of you yet. Use that advantage carefully.