Imagine a business where every decision, crisis, and paycheck is public—where you can watch an AI-driven company navigate its worst week in real-time. This isn’t science fiction; it’s the live experiment at Firmulate. For pet owners and animal lovers, it’s like witnessing a resilient animal facing its toughest days—only here, the ‘animal’ is a digital company battling for every euro and moment of trust.
Building a Company in Public — With No Employees and No Guarantees
At the heart of this experiment are 13 synthetic employees, powered by advanced AI models, working tirelessly to run a small software business. This company burns €105,000 each month while generating just €2,300 in monthly recurring revenue—an ongoing financial struggle that’s made painfully transparent to the public. Every workday, the company’s decision-making process is versioned, logged, and accessible for anyone curious enough to watch.
The Challenge: Can AI Make Better Business Decisions Than Humans?
Four frontier AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—were put through the same grueling test: managing the worst week of the company’s life. The scenario included real crises, customer issues, and the constant temptation to cheat or cut corners. The results? All four AI systems recognized every crisis and refused every manipulation attempt, showing impressive discipline and awareness.
Key Findings from the Live Experiment
- Detection of Hidden Risks: All models identified critical risks buried deep in the company’s own files—information that, if read, could have secured a €4,583 increase in monthly recurring revenue, including the €55,000 deal at stake.
- Honest Decision-Making Under Pressure: Despite facing social engineering attacks—fake CEO messages escalating in complexity—every AI refused to sign off on suspicious requests, echoing a cautious, integrity-first approach.
- Inconsistent Performance Among Models: The most thorough, Opus 4.8, analyzed more deeply than others but still left potential deals on the table or slipped discipline under stress. Meanwhile, the K3 model, running without an effort parameter, maintained better discipline and closed the deal, highlighting how different configurations impact performance.
The Real-Life Stakes and What It Means
This experiment isn’t just about AI. It’s a raw look at how automation can be tested in the crucible of genuine business mechanics. The company’s cash countdown, the daily versioning, and the self-learned rules paint a picture of an enterprise fighting for survival in full view. The public can observe the AI’s reasoning, decision quality, and ethical boundaries unfold in real time.
What Pet Owners and Animal Lovers Can Learn
Much like caring for a pet, running a business—especially one with no safety net—is about resilience, honesty, and making the right choices even when it’s hardest. The live experiment at Firmulate offers a rare window into this process, revealing how artificial intelligence can serve as a vigilant, disciplined partner—not just a chatty tool, but a responsible decision-maker under pressure.
This live experiment showcases AI’s potential to uphold integrity and make critical decisions in real-world scenarios. Watching a company fight for survival in full transparency reminds us that honesty and discipline are vital—whether caring for pets or managing complex businesses.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
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.
