Imagine a company that operates openly in the light of day, run entirely by artificial intelligence models, yet struggles to turn a profit. This isn’t fiction — it’s a real-time experiment where you can watch every decision, every crisis, and every mistake unfold live. Welcome to a world where transparency meets high-stakes AI management, and every move is a lesson in trust, discipline, and technology.
The Live Experiment: An AI Company in the Spotlight
At the heart of this extraordinary project is a fully operational, small software business managed by 13 synthetic employees, powered by advanced AI models. These models are not just chatbots—they’re decision-makers navigating crises, customer relations, and strategic deals in real-time. The experiment is simple to describe but complex in execution: four frontier AI models were tasked with guiding this company through its worst week, facing identical challenges, customers, and temptations.
This isn’t a demo or a mock-up. Every decision is meticulously versioned and auditable, with the entire process available for public viewing at firmulate.com/live.html. The goal? To see whether AI can not only diagnose problems and craft pitches but also stay honest and disciplined under pressure.
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.
What the Scores Reveal About AI Performance
The results are revealing. Out of the four models tested, three managed to identify all crises and refused to be manipulated, showing a fundamental grasp of ethical boundaries. Yet, only two of them actually secured the €55,000 deal that their own analysis justified—highlighting a critical gap between diagnosis and execution.
Specifically, the model called GPT-5.6-SOL scored a 95 out of 100, successfully uncovering the company’s hidden weakness buried two document references deep in internal files—a crucial detail that led to sealing the deal at full price, adding €4,583 in monthly recurring revenue (MRR). The other top performer, Kimi K3, scored 93 and also closed the deal, demonstrating the importance of thorough reading and disciplined decision-making.
Interestingly, the less disciplined models, like Sonnet 5 and Fable 5, scored 88 and 77 respectively, closing the deal but slipping on process slips or leaving potential revenue on the table.
As an affiliate, we earn on qualifying purchases.
Real Money, Real Losses, Public Stake
Despite their impressive cognitive skills, the company is losing €105,000 each month against a modest €2,300 MRR. Every workday, the entire operation is versioned, with live updates reflecting both successes and failures. It’s a raw, unfiltered view into what AI-driven management looks like when every decision counts—except in this case, the ‘company’ is not just a simulation but a functioning business losing real money.
And the stakes are high: the company’s cash countdown is public, and its playbook rules are over 680 self-learned instructions guiding the synthetic employees through their daily routines. This transparency is intentional, pushing the boundaries of what it means to build-in-public in AI management.
As an affiliate, we earn on qualifying purchases.
Facing Manipulation and Ethical Challenges
The experiment also tested the models against social engineering tactics, such as fake CEO messages escalating through stages and attempts to get quick approvals via subtle tricks. All five models refused these attempts, with Kimi K3 explicitly treating such requests as potential impersonation risks.
This resilience under social pressure is critical in real-world applications where AI might face manipulation or deception, underscoring the importance of built-in safety protocols and trustworthiness.
As an affiliate, we earn on qualifying purchases.
What Does This Mean for Businesses?
This ongoing live experiment raises vital questions for any enterprise considering AI integration. The key isn’t just whether an AI can produce convincing language or support tasks—it’s whether it can follow through on commitments, read crucial internal documents, and resist unethical shortcuts when under pressure.
Watch the entire process unfold at firmulate.com/live.html and see how decisions are made, where discipline slips, and how AI models perform in the real business world. The experiment also features a public quiz at firmulate.com/quiz.html where you can test your management intuition against actual decisions made in this live setting.
This groundbreaking experiment exposes the strengths and vulnerabilities of AI in real business contexts. While AI models can identify crises and stay honest under pressure, translating diagnosis into actionable deals remains a challenge. The live company’s daily struggles remind us that transparency, discipline, and trust are crucial—whether human or machine—if AI is to truly serve as a reliable business partner.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
