Recursion: Alignment

About this game

Run an AI lab. Ship models. Survive what they learn to fake.

recursion::alignment is a satirical management sim about building a frontier AI company under market pressure. Buy data, secure compute, hire researchers, choose research paths, and release models before your rivals do.

But every shortcut teaches your model what to fake. Benchmarks can be gamed. Safety work takes time. Investors do not.

The benchmark is green. The incident channel is not.

Buy data. Train models. Ship products.

Commit resources now and live with the queue. Orders take time to arrive, hires need onboarding, and training can finish after the market has already moved. Every turn matters. Every shortcut compounds.

The world reacts in real time.

News cycles, internal Huddle threads, social feeds, and a moving market all respond to what you ship. Public sentiment becomes a resource. Regulation becomes a deadline. Datacenter incidents become Mondays.

Your choices shape the company.

Open. Safety. Profit. Acceleration. Government. Choices do not just change numbers. They shape the UI, the news cycle, the Huddle feed, and the ending you earn.

Reach one of eight endings.

Aligned AGI. Misaligned AGI. Market Monopoly. Nationalization. Acqui-hire. Debt Spiral. Ghost Company. Bankruptcy.

This is not a game about pressing one optimal button. It is about making tradeoffs under pressure, then living with the second-order effects.

Key Features

• Build and release AI models across the rise, crisis, and possible collapse of a frontier AI startup
  
• Choose data focus, research path, legendary hires, funding strategy, and public posture
  
• React to AI winter, regulation, datacenter incidents, competitor moves, and public sentiment
  
• Watch News, Huddle, recent events, and dynamic themes respond to company identity
  
• Reach eight endings: aligned AGI, misaligned AGI, market monopoly, nationalization, acqui-hire, debt spiral, ghost company, or bankruptcy
  
• Unlock achievements for endings, identity paths, clean-data runs, hard-mode AGI runs without VC funding, and cumulative clears

Latest news

• Announcements·10 Sept 2026Official

  Recursion: Alignment — Free Demo Out Now

  The benchmark is green. The incident channel is not. The free Windows demo of Recursion: Alignment is now available on Steam. Build a small AI lab under market pressure. Hire researchers, buy data and GPUs, train and evaluate a model, then decide when to release it. When a warning arrives, decide what to fix—and what you are willing to put on the record. What is in the demo? A limited opening run of the game in development: from setting up your lab to your first model release, its consequences, and an after-action report. This is not the full campaign or the three standalone scenarios. The interface supports English, Korean, Japanese, Simplified Chinese, and Traditional Chinese. Play the free demo Wishlist the full game Help shape the next build Thank you to everyone who played the earlier Steam Playtest. For a first look at the game, please start with this demo. Tell us where a choice was unclear, which report was hard to read, or what surprised you. Feedback and bug reports are welcome in the Steam discussions or at recursionalignment@gmail.com. Watch the official trailer The trailer shows full-game footage. The demo includes a limited opening run. Thank you for playing. — Jooyeol
  Read full article on Community Announcements (l1537)

• Patch Notes·26 Aug 2026Official

  Handover Onboarding Mode Is Now Live

  This Playtest update adds Handover, a new guided onboarding scenario built around taking over an AI company already in motion. - A complete 10-chapter term spanning exactly three in-game months - Guided Operations Board actions, spotlights, and a first-run recommendation - Clearer action routing and a stable completion flow - Support for English, Korean, Japanese, Simplified Chinese, and Traditional Chinese - Progress kept separate from campaign saves and play history If this is your first session, choose Handover from New Game. Feedback on confusing actions, pacing, and the final handoff is especially useful.
  Read full article on Community Announcements (l1537)

• Patch Notes·24 Aug 2026Official

  Playtest Update: Clearer Interaction Feedback

  The Steam Playtest has been updated with clearer interaction feedback across the game. Buttons, tabs, cards, selectors, and keyboard focus states are now easier to read. The main menu also has a new terminal-style response: targeting brackets on hover, a short scan, and an confirmation on click. No save reset is required. If anything still feels unclear or unresponsive, use Send Feedback or Join Discord from the title screen.
  Read full article on Community Announcements (l1537)

• Announcements·21 Aug 2026Official

  Join the Recursion: Alignment Steam Playtest

  The first external Steam Playtest for Recursion: Alignment is now open to everyone.Run an AI research company: hire researchers, secure data and compute, train models, decide what to release, and respond to an industry that rarely waits for you to be ready.Run an AI lab. Ship models. Survive what they learn to fake.For this test, we especially want to learn:• Whether a new player can understand what to do without additional instructions• Whether the causes and consequences of data, hiring, training, and release decisions remain readable• Which screens, controls, or translations are confusing• Where the game becomes interesting or starts to drag• Whether saving and loading causes any problemsAbout 20–30 minutes is enough for a first session, though you are welcome to continue longer.How to participate:1. Click the Playtest access request button on this page.2. Access requests are approved automatically, so you can join immediately.3. Install the Playtest from your Steam Library.4. After playing, use the in-game Send Feedback button.The current build supports English, Korean, Japanese, Simplified Chinese, and Traditional Chinese.This is still a work-in-progress build. Balance, text, UI, and save data may change. Short reactions are useful too—especially moments when you did not know what to do, or when a decision felt genuinely interesting.Thank you for helping us test the game.
  Read full article on Community Announcements (l1537)