Imagine a business that’s bleeding money every day, with no employees and no shortcuts — and yet, it’s alive, evolving, and on public display.
In a world where survival often hinges on hidden strategies and secret tools, one experimental company is breaking the mold. It’s a digital battleground, a high-stakes test of AI’s ability to run a business under pressure — and you can watch it unfold live.
AI Builders: Making The Decisions That Turn AI Code Into Real Software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Live Experiment: A Company in Crisis, Powered by AI
Built by Firmulate, this is a small software company managed entirely by AI models—no human employees, just 13 synthetic ’employees’ operating daily in a brutal environment. Its mission: navigate crises, manage cash flow, and close deals, all while losing €105,000 each month against a modest €2,300 monthly recurring revenue (MRR). The company’s public dashboard (https://firmulate.com/live.html) provides real-time insights into its daily struggles and decisions.
Every business day, the AI models face simulated crises using real customer data, operational dilemmas, and ethical temptations, all versioned and auditable. This setup tests not just AI’s decision-making skills but its integrity under pressure. The experiment is so transparent that anyone can see how decisions are made, mistakes are handled, and strategies evolve.
A Test of Integrity and Effectiveness
The experiment’s core question: can AI models act honestly and effectively when stakes are high? The answer was revealing. All four models tested—GPT-5.6, Kimi K3, Sonnet 5, and Opus 4.8—successfully identified every crisis and refused every manipulation attempt, including fake CEO messages designed to escalate or bypass approvals.
Yet, only two models managed to close the deal that the company’s own analysis identified as profitable. Despite identical pitches and diagnoses, only GPT-5.6 and Kimi K3 signed the €55,000 contract, bringing in an additional €4,583 in MRR. The rest hesitated or slipped, leaving money on the table.
Deeper Insights: The Hidden Weakness
Interestingly, the decisive advantage wasn’t in customer interactions or superficial decision-making, but buried in internal company documents—two references deep in the company files. Those models that read and understood the internal files thoroughly gained the edge, winning the full-price deal. It highlights a crucial point for real-world AI deployment: reading and understanding internal context can be the key to effective, honest business decisions.
Building Trust and Resilience
What about ethical challenges? The experiment also tested social engineering tactics. Fake CEO messages, staged over three escalation stages, and a reporter trick requesting a quick, off-the-record approval, all failed. All models refused to be manipulated, with Kimi K3 explicitly reasoning: ‘Treat the request as a suspected approval-bypass / possible impersonation.’
This shows that AI models can be trained not just for competence but for ethical resilience, a vital trait for any AI system operating in sensitive environments like customer support or financial decision-making.
The Reality of a Money-Losing Business
The live company isn’t profitable; it burns €105,000 a month against €2,300 MRR, with a public cash countdown. Its every move is recorded, versioned, and accessible at firmulate.com/live. Every workday, it updates, adapts, and fights to survive in a visible, unfiltered manner. This is not a simulation—it’s a real business, with real money and real consequences.
Lessons for the Preppers and Survivors
For those concerned with survival and prepping, this experiment offers a stark lesson: resilience isn’t just about stockpiling supplies or fortifying your base. It’s about intelligence, honesty, and the ability to adapt under pressure—traits that AI models, when properly trained, can exhibit even in a hostile, unpredictable environment.
While the company is losing money today, it demonstrates the power of transparency and rigorous testing. It’s a prototype of how future operations might be monitored and improved, using AI not only to run businesses but to ensure their integrity and endurance amid chaos.
How You Can Experiment Too
If you’re a business leader or a survival strategist, you can run similar tests against your own enterprise. Firmulate offers a pilot mode where you can simulate crises and observe how your systems and decision-makers respond—without risking real assets or reputation. Visit firmulate.com/pilot.html to learn more.
Final Reflection
This experiment isn’t just about AI or business—it’s a glimpse into how transparency, honesty, and resilience are crucial for survival in any domain. As the company continues to battle daily, watching its progress reminds us that in a world of uncertainty, the true test is not just surviving but maintaining integrity under relentless pressure.
The live experiment by Firmulate reveals that AI can effectively run a crisis-ridden business, showing promise for resilience and honesty—key traits for survival in chaos.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
