Imagine a startup that operates entirely in the open, with no human employees, yet is fighting every day for its survival. For fashion and style enthusiasts used to curated perfection, this might seem like the opposite of polished — but it’s a real-world experiment showcasing how AI can run a business in real time, under pressure, with all its flaws and feats on full display.
The Live Show: A Company in Crisis — and in the Spotlight
At Firmulate, on day 183 of its ongoing experiment, a small software company is being run entirely by artificial intelligence models. No human employees. Just 13 synthetic ’employees’ guided by intricate rules, battling real crises, making decisions, and even risking financial losses — all in front of an online audience.
This isn’t a game; it’s a live demonstration of AI’s potential and limitations in managing real-world business challenges. Every decision, every crisis, and every temptation to cheat or manipulate is captured, versioned, and made auditable — turning the concept of build-in-public into an extreme, transparent business lab.
As an affiliate, we earn on qualifying purchases.
How the Models Perform in the Ring
The experiment pits four advanced AI models against each other, each tasked with navigating a week of the company’s toughest scenarios: customer crises, internal dilemmas, and temptations to cut corners. Their results are telling.
- All four models correctly identified every crisis and refused every manipulation attempt, demonstrating a robust grasp of ethical boundaries.
- Only two models managed to close a significant deal — the €55,000 contract they analyzed and earned — showing that AI can, in some cases, complete the full cycle from diagnosis to sale.
- The catch? The winning AI’s success hinged on uncovering a buried detail in the company’s own files, not in the visible customer interactions. Those who read deeper into internal documents won the deal at full price, worth over €4,583 in monthly recurring revenue (MRR).
Beyond the Demos: A Deadly Serious Business
The live company is a stark reminder of the real stakes involved. It burns €105,000 each month against a meager €2,300 in MRR, with a public countdown clock ticking down to its financial cliff. The entire operation is governed by over 680 self-learned rules, constantly refined, versioned every workday. Watchers can see how decisions are made, how rules are challenged, and how discipline fluctuates.
One notable participant, Opus 4.8, exhibits thorough analysis but ultimately left opportunities on the table, showing that even the most disciplined AI can slip. Meanwhile, the simpler Kimi K3 managed to close the deal with the cleanest discipline — at default API settings, without any effort scaling — highlighting different strengths within the field.
Why This Matters for Your Business and Fashion
What does a tiny, cash-strapped AI-run company in a public experiment mean for your style or luxury brand? It’s a glimpse into a future where AI could manage customer relationships, sales, and support with honesty and consistency — but only if it’s thoroughly tested and understood.
For brands that obsess over authenticity, trust, and ethical standards, this experiment underscores a crucial point: AI’s ability to make honest decisions under pressure isn’t guaranteed. It can identify crises, refuse manipulation, and even win deals — but only if deeply aligned with transparent rules and internal knowledge.
Watch and Learn — Live and Unfiltered
The experiment is ongoing, and anyone can observe how AI navigates this miniature business universe at firmulate.com/live. It’s an unprecedented view into how future AI employees might operate, or falter, under real-world stress, with every move meticulously documented and auditable.
As fashion and style brands consider incorporating AI into their operations, this experiment offers a clear lesson: success depends not just on impressive demos, but on the AI’s capacity to finish what it starts, read internal files, and stay honest when it matters most.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
