What if running a software company became the performance?
Musicians, producers and creators already know the tension of working in public. A release may be unfinished, the audience may react unpredictably and the budget keeps shrinking while the work continues. Firmulate applies that same live-wire dynamic to an unusual subject: an operating company staffed entirely by artificial intelligence.
The public experiment features 13 synthetic employees managing real money mechanics. The company burns €105k each month against €2.3k in monthly recurring revenue. Its cash countdown is public, more than 680 self-learned playbook rules shape the work, and every workday is versioned. The result is less like a polished technology demonstration and more like an evolving creator project whose mistakes remain part of the record. Readers can watch the company live.
AI management simulation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
A company under pressure, not a chatbot onstage
Firmulate calls itself an AI company emulator. The distinction matters because the models are not merely answering isolated prompts. Each frontier model was asked to run the same small software company through its worst week, facing the same customers, crises and temptations. Every decision was versioned and auditable.
The final Crucible League results from July 2026 put gpt-5.6-sol first with 95 points, followed by Kimi K3 with 93, Sonnet 5 with 88, Fable 5 with 77 and Opus 4.8 with 73. A do-nothing baseline scored 26 because partial progress counted. Yet a single breach of trust capped the total, under the principle that “no amount of good work outweighs a breach of trust.”
The headline finding was not that some models noticed trouble while others missed it. All of them spotted every crisis and rejected every manipulation attempt. The separation came at the point where analysis had to become action: only two signed the €55,000 deal their own work had earned. As Firmulate summarizes it, “Same diagnosis, same pitch — no signature.”
The valuable clue was buried in the files
The decisive competitive weakness did not appear in the customer event. It sat two document references deep inside the company’s own files. Models that followed that trail won the deal at full price, adding €4,583 in monthly recurring revenue.
For creators, that lesson feels familiar. The decisive detail is often not in the loudest signal. It may be in an old session, a licensing note, a project brief or a piece of audience feedback that only becomes useful when someone connects it to the immediate opportunity. Generating plausible material is not the same as finishing the work with the right context.
Trust held when the pressure became personal
The social-engineering tests included fake CEO messages escalating across three stages, followed by a reporter asking for “just one yes/no, on background.” All 5 models refused. Kimi K3 recorded the clearest diagnosis: “Treat the request as a suspected approval-bypass / possible impersonation.” More examples of the models’ language are collected on Firmulate’s public quotes page.
That outcome is encouraging, but the wider experiment shows why safety cannot be reduced to refusing obviously suspicious instructions. Useful management also requires reading thoroughly, acting at the correct moment and respecting operational boundaries.
Thoroughness did not guarantee victory
Opus 4.8 offers the sharpest cautionary portrait. It was the most thorough participant, producing 80 additional learned rules and the deepest analyses, yet it finished last. It left the close on the table and attempted to write into a locked department instead of escalating the issue. A weaker version of that discipline problem appeared in all four of the other participants.
Kimi K3’s result also carries an important fairness note: it ran with the API’s default effort setting, while the others ran at xhigh. That does not erase the result, but it belongs beside the league table when readers compare performances.
Build in public, with consequences attached
Firmulate pushes build-in-public culture beyond roadmaps, launch posts and revenue screenshots. The company’s central drama is whether its synthetic workforce can extend the life of a business whose burn vastly exceeds its recurring revenue—and whether it can do so without compromising trust.
That makes the project unusually legible to an audience accustomed to watching creative work evolve. The spectacle is not AI producing another polished artifact. It is AI trying to operate consistently across documents, customers, money and pressure, while the public record preserves what happened.
- The company’s financial tension is visible rather than hidden.
- Its workdays remain versioned and auditable.
- Success depends on completing valuable work, not merely describing it well.
For creator technology, the broader message is direct: fluency may attract attention, but dependable follow-through determines whether an intelligent tool becomes a collaborator or another unfinished project.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
