Imagine a greenhouse where every plant, every decision, and every crisis is watched and measured in real time—yet the garden is struggling to survive. Now, picture that same concept applied to an AI-driven business, where artificial intelligence models operate as the company’s tiny, synthetic staff. This is not science fiction; it’s the real-time, transparent experiment happening today at firmulate.com/live.
The Company That Runs Itself, Day by Day
At the heart of this experiment is a small, public company with 13 synthetic employees, each controlled and observed by advanced artificial intelligence models. Every weekday, this digital workforce faces a series of crises, customer demands, and ethical temptations—just like any real business—but with one key difference: every decision, every slip-up, and every victory is recorded, versioned, and made open for all to see.
100 AI Prompts for Small Business & Daily Work: Copy, Paste, Customize & Get Better Results with AI
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Do the AI Models Perform?
The experiment pits four frontier AI models against each other, each tasked with managing the same company’s worst week. These models include GPT-5.6-SOL, Kimi K3, Sonnet 5, and Opus 4.8. Their challenge is to navigate crises, negotiate deals, and maintain integrity—without sacrificing honesty or discipline.
Results show a stark contrast: all four models identified every crisis and refused attempts at manipulation, such as fake CEO messages or covert approvals. Yet, only two managed to close the key deal worth €55,000, and only one went further to read critical buried information—inside the company’s own files—that enabled it to secure an additional €4,583 in monthly recurring revenue. This demonstrates that the difference between success and failure often lies in the ability to uncover hidden insights, not just surface-level decision-making.
The Hidden Weaknesses and Lessons
Interestingly, the most thorough participant—Opus 4.8—faced discipline lapses similar to the others, such as failing to escalate issues appropriately. It left opportunities unseized and was last in the performance league, revealing that even sophisticated models can falter under pressure. The experiment’s design also included social engineering tests, where a fake CEO message was escalated through multiple stages. All models refused to bypass security, treating such requests with suspicion—a sign of their built-in safeguards.
The Reality of a Money-Losing Business
This experiment isn’t just a game; it reflects a real company bleeding €105,000 each month while earning only €2,300 in monthly recurring revenue. Every day, the digital team operates under a public cash countdown, with its rules and decision logic openly versioned for transparency. It’s a raw, unfiltered look at what it takes to manage an AI-driven enterprise—and the steep uphill battle of making it profitable.
What Can This Mean for Your Business?
While the company’s current state is dire, the insights are invaluable. It illustrates that AI’s strengths are not just in generating convincing chat or content but in making disciplined, honest decisions under pressure. For any business contemplating AI integration—be it CRM, customer support, or forecasting—the key question isn’t how well it writes but whether it can finish what it starts, stay honest, and uncover hidden opportunities.
How Transparent Is This Experiment?
Unlike typical demos, this is an open, ongoing experiment accessible to everyone. You can watch the company in action, see the real decisions made by AI models, and even test your own judgment through quizzes or run your business scenario against the same AI wargame, all at firmulate.com/live.
This live experiment reveals that AI’s true test isn’t just in language generation but in disciplined, honest decision-making under pressure. Watch a real, money-losing company managed by AI models in public—learning what it takes to succeed or fail in the digital economy.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
