AIThis post was created with the assistance of artificial intelligence (AI).

For readers tracking how AI systems might coordinate complex work, GewerkTon offers a bounded case to examine: AI Company Emulator presents an emulation of AI agents advancing a construction-site app, with each decision available to inspect. The replay begins from the startup’s real state, but customers, deals, revenue and figures after day zero are simulated; they are not real business results.

Prime Big Deal Days · Oct 6–7Offer from Amazon

Get the latest gadgets delivered free — and shop member deals

  • Fast, free delivery on millions of items
  • Access to Prime Big Deal Days deals on October 6–7
  • Prime Video, Amazon Music and more included
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

How it works: Each role in the company is its own AI agent — product, engineering, pilot success, business development and finance. The world around them is simulated too: prospects answer (or don’t), pilots send feature requests with deadlines, and customers get impatient when promises slip. The team notices, decides, acts and learns; every business day is a git commit that the replay walks through. A live feed shows each decision, an office map shows who is working, and a timeline lets you jump to any day.

Simulated replay running from day 38 to day 44: the counters change while the emulated AI employees work and post decisions in the live feed
The replay in motion: days 38 to 44 of the emulated run. Everything after day 0 is simulated.

The starting point is real: the run starts from GewerkTon’s actual state — one founder, an experienced site manager testing the app in beta, and no customers. Everything after day 0 is emulated, and the page says so on every screen.

What the replay shows: in the current run (days 1–44, as of 30 September 2026) the team wins its first pilot on day 6, ships its first requested feature on day 16, and on day 44 turns a pilot into the first paid licence. Along the way it gets stuck — reviews that reject everything, offers that never get recorded — and the founder steps in with short directives that you can read in the feed. That is the point: you see not just the wins but where an AI team stalls and what unblocks it.

Diagram of the simulated replay: GewerkTon's real starting state goes into the AI Company Emulator; simulated releases, pilots and a live feed come out
What goes in is GewerkTon’s real starting state; everything that comes out is simulated.

What it isn’t: a customer case study. The companies, deals and figures after day 0 are simulated, not real GewerkTon traction — the replay is an honest look at how an AI workforce behaves when it runs a real product’s business.

The emulator behind it is Firmulate (firmulate.com), which runs AI models as complete companies — with crises, money mechanics and temptations — and scores management quality, not chat quality. Watch the GewerkTon replay at aicompanyemulator.com.

FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

The Lost Treasure Of Sid Meier’s Pirates

A hidden in-game treasure called ‘The Lost Treasure’ has been uncovered in Sid Meier’s Pirates, prompting excitement among players and historians.

Reproducible AI Benchmarks: The Role Of UK AISI And EvalEval

UK AISI is sharing selected benchmark results through EvalEval Evaluation Cards, with verification, evaluation context and configuration details.

Where The 176GB Of AI Memory Is Hidden — The Critical Details

An in-depth analysis of how AI memory is allocated, revealing that model weights are only part of the story. The KV cache and other factors often cause unexpected memory issues.

Show HN: A Project Oberon System Version Running On RISC-V Instead Of RISC-5

A developer has shared a version of the Project Oberon operating system ported to RISC-V architecture, replacing the original RISC-5 processor support.