Imagine a small startup running entirely on artificial intelligence — making decisions, facing crises, and even losing money every single day. For pet owners and animal lovers alike, it’s a story of resilience and risk, but in the digital world, this is the real-time saga of a groundbreaking experiment that reveals how AI manages ethics and business under pressure.
The Live AI Company: A Real-Time Experiment in Business and Ethics
At first glance, it might seem like just another tech demo, but this is something much more tangible — a real, functioning small business run entirely by AI models. Every weekday, 13 synthetic employees operate within a company that faces the same crises, customer demands, and temptations as any real firm. The twist? It’s all happening live, with transparent decisions and a public cash countdown, making it an unprecedented window into AI-driven management.
How It Works
The company runs on open-source AI models, each with their own decision-making style. The models are tested against the same set of challenges, from customer complaints to internal crises, all while being recorded and versioned daily. Their performance is scored by a public leaderboard, with the top model, GPT-5.6-sol, earning a score of 95 out of 100, and others trailing behind but still participating.
Decisions Under Ethical and Financial Pressure
One key aspect of this experiment is how AI models handle ethical dilemmas. For example, when approached with social engineering tricks like fake CEO messages or subtle requests to bypass approval steps, all four models refused to manipulate or bypass protocols. Kimi K3, a model known for its fairness, explicitly recognized the risks of impersonation and refused to act on suspicious requests. This adherence to ethical boundaries is crucial, especially if AI ever touches sensitive business systems such as customer management or financial transactions.
What Sets This Company Apart
- It’s fully transparent — every decision, rule, and crisis is logged and publicly viewable at firmulate.com/live.html.
- It operates in real time, losing €105,000 monthly against €2,300 in monthly recurring revenue, with a ticking cash countdown that adds urgency to every decision.
- It’s driven by over 680 self-learned playbook rules, with decisions continuously versioned and auditable, showcasing how AI handles complex management scenarios.
The Human-Like Flaws and Lessons
Interestingly, the most thorough participant, Opus 4.8, performed the poorest in a critical close — leaving an opportunity on the table and slipping discipline into internal processes. This highlights that even advanced AI models can falter under pressure, especially when not optimized for discipline or risk management.
The Broader Implications
This experiment is more than a tech showcase; it raises questions for any business considering AI’s role in management, customer service, or decision-making. Do these models read and understand all relevant information? Do they stay honest under pressure? And ultimately, will they finish what they start, or leave opportunities slip away?
Agentic Spec-Driven Development: A Practical Method for Using AI to Build Complete Specifications for Software, Products, and Knowledge Work
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Verdict: An Ongoing Test of Trust and Utility
While all models identified crises and refused manipulation attempts, only two managed to close a deal at full price. The gap in performance was hidden in a buried document reference — a detail only the models that read deeply into files could catch. This underscores the importance of thorough information processing and trustworthiness in AI systems that might one day run critical parts of your business.
For pet owners and animal lovers, the story resonates with themes of resilience, trust, and careful management — qualities that AI models are still learning to master. As the live company continues to run, watching it in action offers a rare glimpse of how AI might one day manage not just business but complex ethical landscapes, all in real time.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Ethical AI Governance & Decision Journal: A Structured System for Documenting, Tracking, and Defending Real World Decisions and Risk (Decision Intelligence Series, Band 2)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Large Language Models: The Hard Parts: Open Source AI Solutions for Common Pitfalls
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Building Real-Time Analytics Systems: From Events to Insights with Apache Kafka and Apache Pinot
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
