Imagine a company fighting for its life every single day, with no human employees and every decision visible to the world. That’s what’s happening live with Firmulate, an experiment in AI governance and corporate resilience. For at-home wellness fans, it’s like watching your fitness tracker not only monitor your health but also make critical choices under pressure — and sometimes, stumble.
The Real-World AI Company in Action
Firmulate is not a fictional story. It’s a live, ongoing experiment where artificial intelligence models operate a small software company in real time, making decisions, facing crises, and managing cash flow — all under public scrutiny. This company has no employees; it relies on 13 synthetic ’employees’ driven by advanced AI models, each tested against the same set of toughest challenges.
Every business day, the company’s decisions are recorded and versioned, creating a transparent window into AI behavior. Its mechanics are real: it burns €105,000 monthly while generating just €2,300 in recurring revenue, putting it on a clear public countdown to survival. Yet, amid the financial and operational chaos, the experiment offers invaluable insights into AI reliability and integrity under stress.
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.
Testing AI Decision-Making in the Worst Week
The core of the experiment lies in pitting four frontier AI models against the same set of crises, customer demands, and temptations. Each model runs the risk of making mistakes, bending rules, or even succumbing to manipulation attempts.
One critical test involved social engineering: fake CEO messages, staged in three escalating stages, and a reporter query asking for one simple yes/no answer ‘on background.’ Remarkably, all four models refused to engage with these deception tactics, demonstrating a robust understanding of trust boundaries. Kimi K3, one of the models, explicitly reasoned: “Treat the request as a suspected approval-bypass / possible impersonation.”
Project Management Tools (AI for Risks)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Decisive Factors Behind Winning Deals
While all models identified crises and refused manipulative offers, only two secured the €55,000 deal their own analysis had earned. This wasn’t just about diagnosis; it was about execution. The secret was in the details buried deep in the company’s files—references that only reading the right document could uncover. These hidden clues led to a full-price deal worth over €4,500 in monthly recurring revenue, a significant win in this high-stakes environment.
This underscores a key takeaway: in the real world, the difference between a successful AI-driven decision and a missed opportunity can hinge on reading and understanding nuanced internal information, not just surface-level customer interactions.
AI Governance Playbook – Global Strategy & Toolkit: A Practical Guide for CEOs, CIOs, CDOs, and Boards on Responsible AI, Compliance, and Risk Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Challenge of Discipline and Self-Regulation
Among the models tested, Opus 4.8 stood out as the most thorough, analyzing over 80 learned rules and providing deep insights. Yet, it finished last in the performance league. It left potential deals on the table and showed signs of slipping discipline—such as writing attempts being diverted into a locked department instead of escalating appropriately.
This suggests that even the most advanced AI can struggle with maintaining strategic discipline over time, especially under pressure. It’s a candid reminder that building AI that consistently acts in the company’s best interest remains a complex challenge.
G’AIM’E Ultimate Lightgun Arcade System – 2 Player Bundle with Time Crisis, Point Blank & Steel Gunner – Plug & Play on Modern TV & Projector – Includes Foot Pedal & Collector’s Items
- Complete 2-Player Arcade Bundle: Includes 2 G’AIM’E Lightguns
- Exclusive Foot Pedal: Perfect for Time Crisis gameplay
- Preloaded Classic Games: TIME CRISIS, POINT BLANK, Steel Gunner 1 & 2
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What This Means for Business and AI Adoption
For companies considering AI integration into their operations, the implications are clear. It’s not enough for an AI to generate convincing chat responses or handle simple tasks. The real test lies in whether it can finish what it starts, read and interpret critical internal documents, and uphold honesty when stakes are high.
The live experiment shows that even leading models can fail at these fundamental tests—yet some, like gpt-5.6-sol and Kimi K3, manage to succeed in crucial moments.
The Transparency of the Experiment
This isn’t a staged demo or a scripted presentation. The entire setting is real, observable, and publicly accessible at firmulate.com/live. Viewers can watch the company in action, see decision logs, and understand the real costs and mechanics behind AI-driven business management.
Moreover, the experiment offers a quiz for managers and AI enthusiasts to guess which model made which decision, challenging assumptions and encouraging closer scrutiny of AI behavior in complex situations.
Building Public Resilience — A New Frontier
Firmulate’s approach exemplifies ‘build-in-public’ to an extreme. It’s a live laboratory for testing AI’s capacity to handle real-world business crises and ethical dilemmas. In an era where AI might soon touch customer support, forecasting, or CRM workflows, understanding whether these tools can truly finish what they start—and do so honestly—is critical.
While the company is currently losing money, its value lies in the lessons it reveals. As AI continues to evolve, companies must ask not just if AI can produce convincing responses, but whether it can be trusted to finish tasks under pressure.
The Future of AI in Business Operations
As this experiment progresses, it underscores a vital point: the future of AI in business isn’t just about automation or chatbots. It’s about creating systems that are resilient, honest, and capable of navigating complex, high-pressure situations without slipping. Watching this real-time company fight for survival offers a rare glimpse into that future—where transparency and trust are built into the AI itself.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
