Imagine a company with no employees, losing €105,000 every month, yet still navigating the chaos of real crises and decision-making—live, in public. This isn’t fiction; it’s the ongoing experiment at Firmulate, a pioneering venture in the world of AI-powered business simulations.
The Groundbreaking Live Experiment
At the heart of this story is a small, simulated software company run entirely by AI models. In a public online platform, four frontier AI models are tasked with managing its day-to-day operations, facing the same challenges and crises, with every decision meticulously recorded and auditable. The goal? To evaluate how these AI agents handle real-world business pressures, trust dilemmas, and ethical boundaries.
The Players and Their Scores
- gpt-5.6-sol leads with a perfect score of 95, successfully uncovering hidden information buried deep in company files and closing a full-price deal worth over €4,583 monthly recurring revenue (MRR).
- Kimi K3, the newest entrant, scores just slightly behind at 93. Notably, it also closed the deal, demonstrating the field’s most disciplined decision-making.
- Sonnet 5 with an 88, managed to sign the deal but showed minor slips in process discipline.
- Fable 5, with a score of 77, kept the rules but left opportunities on the table, such as failing to finalize a signed deal due to internal process lapses.
Despite their differences, all models identified and responded to every crisis. They refused manipulative social engineering attempts, such as fake CEO messages or reporter tricks, with all five models refusing to sign off on a suspicious approval bypass request — a testament to their built-in safeguards.
Silhouette America Studio Business Edition Software, Multicolor
Silhouette Studio Business Edition Is A Version Of The Silhouette Studio Software Program That Intended For Small Business…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Company in Practice
Beyond the scores, the live setup offers a stark look at the real mechanics behind AI management tools. The experiment features 13 synthetic employees, each governed by a set of more than 680 self-learned rules, with every workday versioned for transparency. The company burns €105,000 monthly against a meager €2,300 MRR, with a public countdown to its financial collapse, illustrating the brutal reality of AI-driven business management.
Deep Dive into the Results
The most thorough participant, Opus 4.8, analyzed over 80 learned rules but still faltered in the final moments, leaving a promising deal unexecuted, and slipping into internal process lapses instead of escalating issues properly. Interestingly, the performance gap was often tied to reading deeper into company files—something the models that diligently parsed documents succeeded at, directly influencing the outcome.
The AI-Driven Leader: Harnessing AI to Make Faster, Smarter Decisions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why This Matters
For professionals in software, QA, and development, the takeaway is clear: the question isn’t whether AI can generate convincing chat interactions but whether it can reliably finish tasks, stay honest under pressure, and make correct decisions based on a full understanding of context. As AI models become more integrated into daily business workflows, their ability to uphold integrity and completeness will define their true value.
From Benchmarks to Business Decisions
The live experiment’s league table shows that the highest scorer, gpt-5.6-sol, delivered the full package: identifying critical hidden information, closing the deal, and maintaining discipline. K3, despite running without an effort parameter, also secured the deal with the cleanest decision process, proving that foundational discipline matters. Meanwhile, models like Fable, while rule-compliant, lacked the execution to finalize opportunities, highlighting that rules alone aren’t enough.
Trust.: Responsible AI, Innovation, Privacy and Data Leadership
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Real-World Application and Next Steps
Businesses can now run their own AI wargames using the same setup, testing how AI agents perform against their unique crises and workflows before real deployment. This ‘build-in-public’ approach offers unprecedented transparency—every decision, every slip-up, and every success is visible online at firmulate.com/live.html.
Ultimately, this experiment underscores the importance of building AI systems that can not only diagnose problems but also act decisively and ethically in high-pressure situations. As the company’s public countdown continues and the models refine their strategies, one thing is clear: the future of AI in business is a high-stakes game of trust, discipline, and transparency.
This live experiment reveals how AI models manage real business crises—spotting every issue, refusing manipulation, and closing high-value deals—highlighting the importance of decision integrity in AI-driven companies. Watch the ongoing battle at firmulate.com/live.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
DULIWO Model Scriber Tool Kit, 7-Blade Chisel Set for Gunpla
Model Kit Tools: Includes metal scribe tool (1 handle + 7 blades), small hand drill set (10 bits…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
