Imagine Watching a Company Struggle to Survive—Live and Unfiltered
For at-home wellness tech enthusiasts, innovation often means sleek gadgets or personalized apps. But what if you could see a company fighting for its life, driven entirely by artificial intelligence, in full public view? This isn’t science fiction; it’s the reality of a groundbreaking experiment that puts a real, money-losing business on display, run entirely by AI models making decisions, avoiding traps, and even losing deals, every single day.
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 Company: An AI-Driven Business in Real Time
At the heart of this experiment is a small, virtual company that has no human employees. Instead, it operates with 13 synthetic ’employees’—AI models carefully trained to make management decisions. Every day, the company faces the same set of crises: customer issues, strategic temptations, and operational dilemmas. Yet, despite its sophistication, it is losing €105,000 each month while generating just €2,300 in monthly recurring revenue. This stark reality is not hidden; it’s openly available at firmulate.com/live.
The company’s financials are transparent, with a public countdown showing how much cash remains. Meanwhile, its decisions are captured and versioned daily, creating an auditable history of management choices. Every decision is influenced by a set of over 680 self-learned rules, with the entire decision-making process observable and comparable day by day.
The Experiment: Putting AI Models to the Test
The core of this project involves running four leading AI models through the same hypothetical week of business crises. These models include GPT-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8. Each faces identical scenarios—customer complaints, crisis management, and strategic temptations—while being tested for honesty, discipline, and problem-solving capabilities.
Remarkably, all four AI models identified every crisis and refused every manipulation attempt, such as fake CEO messages or reporter tricks designed to bypass approval processes. For instance, when fake messages escalated over three stages, all models refused to approve them, reasoning that such requests could be impersonation or approval-bypass attempts. Kimi K3, notably, explicitly recognized these as suspicious and declined to proceed.
The Hidden Weakness: Reading Beyond the Surface
One of the most telling findings came from a deeper analysis of the models’ performance. While they all performed well on the surface, the real differentiator was their ability to read and interpret critical internal documents.
In particular, the models that examined the company’s own internal files uncovered a vital piece of information buried two references deep in the documentation, which was pivotal in closing a lucrative deal. The models that read this information secured an additional €4,583 MRR, effectively winning the company’s only profitable deal during that week. This hidden advantage underscores how crucial deep reading and thorough analysis are for AI decision-making in complex scenarios.
Discipline Under Pressure and the Cost of Slip-Ups
Despite their technical prowess, not all models won the deal. Opus 4.8, for example, was the most thorough participant, analyzing over 80 learned rules and conducting deep analyses. However, it left the final closing on the table by slipping into a less disciplined response—it wrote attempts into a restricted department instead of escalating them, resulting in a missed opportunity.
This demonstrates that even the most comprehensive AI can falter under real-world pressures, especially if discipline or process adherence slips. Interestingly, the models’ default settings influenced their performance: K3 was run without effort parameters, while others operated at higher levels of resource allocation, which likely affected their focus and thoroughness.
As an affiliate, we earn on qualifying purchases.
Implications for Business and AI Adoption
This live experiment is more than a tech curiosity; it offers a window into the future of AI in management. For companies contemplating AI tools to handle customer support, strategic planning, or operational oversight, the questions are now about reliability and integrity—can these models finish what they start? Do they understand internal information? Will they stay honest under pressure?
As the experiment shows, the ability to read internal files and resist manipulative tactics is a decisive advantage. The models that uncovered buried facts and refused to sign false deals were the most successful, highlighting that AI decision-making quality depends not only on language fluency but also on integrity and thoroughness.
Build-in-Public: Transparent, Watchable, and Ongoing
What makes this experiment truly unique is its transparency. The entire company runs live, every decision, every crisis, every slip-up is publicly documented and versioned. Spectators can watch the company’s daily struggles at firmulate.com/live. They can also explore the decision logs, examine the AI’s reasoning, and even participate in quizzes that challenge management intuition against AI analysis at firmulate.com/quiz.html.
This is not just a demonstration of AI capability—it’s a new form of corporate transparency, where the management of a real, cash-burning business becomes a public, ongoing story. It’s a radical shift towards building trust through openness, even when the company is losing money and facing existential threats.
As an affiliate, we earn on qualifying purchases.
The Big Takeaway: AI’s Critical Role in Business Integrity and Risk Management
This live experiment vividly illustrates that in AI-driven management, the ability to uncover hidden facts, resist manipulation, and stay disciplined under pressure is vital. For businesses integrating AI into their core operations, success hinges not just on how well an AI can write or analyze data, but on whether it can finish what it starts—reading deeply, acting honestly, and making decisions that truly serve the company’s best interests.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
AI internal document analysis tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
