Watch an AI-Run Company Fight for Survival in Real Time — No Employees, Just Rules and Crises

Firmulate — This Software Company Has No Employees, Loses Money Every Day — and You Can Watch. — Live on firmulate.com.

Imagine a small startup, without a single human employee, battling daily crises, making tough decisions, and trying to stay afloat—all watched live by the world. What if you could see this company struggle, cheat, fail, and sometimes succeed, all in real time? Welcome to the most extreme experiment in building AI management systems—where transparency becomes a test of trust, discipline, and technology.

The Live Experiment: A Company Without Humans

At the heart of this groundbreaking venture is a live, publicly accessible platform that runs an entirely synthetic company—no human staff, just AI models acting as decision-makers and managers. The company is managing real money mechanics: it burns €105,000 each month against a revenue of only €2,300. Every decision taken by the AI models is recorded, versioned, and open for scrutiny at firmulate.com/live.html.

Express Schedule Free Employee Scheduling Software [PC/Mac Download]

User-friendly drag & drop interface: Simple shift planning
Manage time-off and holidays: Add sick leave, breaks, holidays
Email schedules to employees: Send schedules directly via email

View Latest Price

As an affiliate, we earn on qualifying purchases.

How the AI Models Are Tested

Four advanced AI models, based on frontier language AI, are given the same set of challenges: a week of crises, customer demands, and temptations to cheat. Each model approaches the same scenario, and their decisions are compared. The goal? To see whether these models can identify problems, resist manipulation, and close deals—just like human managers should.

Results That Speak for Themselves

All four models spotted every crisis and refused every manipulation attempt, showing a remarkable level of discipline.
Only two of the models managed to close a €55,000 deal, matching their own analysis and pitch—an essential measure of operational efficacy.
Interestingly, the decisive advantage was in reading deeper into the company’s own files—hidden references that the other models missed. The model that found the buried fact closed a deal worth over €4,500 in monthly recurring revenue.

Beyond crises and opportunities, the experiment tested how the models respond to social manipulation. Fake CEO messages and a reporter trick were thrown into the mix. All models refused to be duped, with Kimi K3 reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.”

The Real Money Mechanics and Challenges

The experiment isn’t just about AI performance; it exposes the brutal reality of managing a company with AI. The live setup reveals a daily cash countdown, with the company burning €105,000 each month against minimal income—€2,300. The platform tracks over 680 self-learned rules, with every decision and process versioned, providing transparency and a detailed view of AI decision-making in crisis mode.

Insights from the Latest Results

The most thorough participant, Opus 4.8, learned over 80 rules and performed deep analyses but still left a deal unclosed and slipped in discipline—highlighting that even the most disciplined AI can falter without the right parameters. Meanwhile, Kimi K3 ran without an effort parameter—using default settings—and still managed to close the deal, showcasing how different configurations affect results.

Why This Matters for Business and AI

This experiment underscores a critical point: the real challenge for AI in management isn’t just generating convincing chat or reports. It’s about how well AI can finish what it starts, stay honest under pressure, and read the full context—like uncovering hidden clues in internal files that can make or break a deal. It’s a stark reminder that AI must be tested against real-world crises, not just in idealized demos.

Experience It Yourself

Interested in seeing this experiment in action? You can watch the ongoing weekly runs, examine the live decision logs, and even run your own business wargame against a read-only export of this virtual company at firmulate.com/quiz.html and firmulate.com/pilot.html. This is a transparent, built-in-public showcase of how AI management models behave under stress—a true test of trust, discipline, and utility.