Watch a Company Run by AIs Struggle to Survive — and You Can See It Live

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

AUDIBLE

Listen free for 30 days with Audible

Thousands of audiobooks and originals — cancel anytime.

As an affiliate, we earn on qualifying purchases.

What if the kitchen never stops fighting to stay open?

Imagine a restaurant where every decision, from ordering ingredients to handling a customer complaint, is made by artificial intelligence. Now, picture that same restaurant losing money every single day, yet remaining open for you to watch in real time. This is not science fiction — it’s the live experiment of a software company run entirely by AI models, now streaming online.

AI Builders: Making The Decisions That Turn AI Code Into Real Software

View Latest Price

As an affiliate, we earn on qualifying purchases.

The Reality of a Living AI-Managed Business

This groundbreaking experiment, hosted by Firmulate, presents a small software company operated by 13 synthetic employees — AI models that simulate decision-making in a real business environment. The company is in constant crisis: it burns €105,000 each month while generating only €2,300 in monthly recurring revenue. Yet, every workday, its decisions, struggles, and even mistakes are publicly available at firmulate.com/live.html.

How Does It Work?

The company faces the same challenges every day: customer demands, crises, and temptations to cheat or manipulate. Four advanced AI models, each representing a different approach, are tasked with running this miniature business through its worst week. Every decision they make is versioned and transparent, allowing viewers to see what choices they consider and why.

The Results So Far

Remarkably, all four models detected every crisis and refused every manipulation attempt, such as social engineering tactics. Only two managed to secure the €55,000 deal that their own analysis told them was possible — a sign that the models are learning to recognize opportunities but are still conservative about closing deals in complex situations.

One of the models, Opus 4.8, displayed the deepest analysis with over 80 learned rules but still finished last because it left some opportunities on the table and slipped on discipline, preferring to write attempts into a locked department rather than escalate. This shows that even the most thorough AI can struggle with human-like discipline and organizational process.

What’s the Hidden Weakness?

Interestingly, the models that successfully closed the deal did so by reading deeper into the company’s own files, uncovering a critical piece of information buried two documents deep. This hidden insight made all the difference, earning an additional €4,583 in monthly recurring revenue.

The experiment also tested how the AI models handle social engineering — fake CEO messages, staged reporter requests, and impersonation attempts. All five models refused to be manipulated, adhering to built-in safeguards. Kimi K3, for example, explicitly treated such requests as potential impersonation, demonstrating the importance of trust and security in AI decision-making.

The Real Money and the Public Live Experiment

The company’s entire operation is visible online, with real money at stake. It burns €105,000 each month against a tiny stream of €2,300 in recurring income, illustrating the stark reality of AI-managed business — it’s not about perfect chat, but about real decision quality under pressure.

This experiment is accessible at firmulate.com/live.html. It is a raw, unfiltered glimpse into how AI models handle crises, ethics, and profit in a real-world scenario. Viewers can watch daily decisions unfold, see the decisions that lead to success or failure, and even guess which model made each choice through a live quiz at firmulate.com/quiz.html.

Why Does This Matter?

For businesses considering AI automation, the key question isn’t whether the AI can write a good paragraph — it’s whether it can get things done, stay honest, and finish what it starts. This experiment exposes the strengths and flaws of current AI decision-making in a context that closely resembles real business operations, not just chatbots or demos.

Conclusion: A Glimpse into the Future of AI Workforces

This ongoing, transparent experiment offers a rare window into how artificial intelligence might run (or fail to run) real companies. It underscores the importance of trust, discipline, and the ability to uncover hidden insights, all critical factors for AI to succeed in the workplace. As the company continues to burn cash and battle for survival — live and in public — it provides a compelling case study for anyone interested in the future of AI and automation.