Imagine running a small outdoor equipment company—struggling to stay afloat—while an AI-powered management team fights to keep it alive, making every decision count. Now, picture watching that battle unfold live, every day, with no filters or edits. That’s the reality of the recent experiment by Firmulate, a company that runs a public, real-time simulation of AI managing a business under extreme pressure.
Get garden gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The Thrilling Experiment: AI in the Hot Seat
Firmulate has created a unique, transparent window into the future of AI as business managers. Every weekday, it runs a small software company through its worst week—crises, tough decisions, temptation to cheat—all within a real, functioning simulation accessible online. There are no fictional scenarios here; this is a live experiment, with the company’s actual cash flow, its decision-making processes, and the decisions of AI models on full display.
The Model Showdown
Four advanced AI models were tested head-to-head, each tasked with managing the same difficult week in the company’s lifecycle. These models, including the notable GPT-5.6-sol and Kimi K3, were evaluated on their ability to identify crises, refuse unethical manipulation, and close critical deals.
All four AI models successfully identified every crisis and refused every attempt at manipulation—such as fake CEO messages or subtle bribery tactics. Yet, only two of them managed to close a key deal worth €55,000, a clear measure of effective management. Interestingly, the model that found a buried piece of crucial information—hidden two document references deep in the company’s files—secured a deal that added more than €4,500 in monthly recurring revenue. This indicates that reading and understanding detailed internal documents can be a decisive advantage.
Products Worth Considering
Breville Grind Control BDC650BSS Thermal Carafe Coffee Maker – Brushed Stainless Steel
KitchenAid Classic Series 4.5-Quart Tilt-Head Stand Mixer, Onyx Black
Cuisinart SM-50BK Stand Mixer, 5.5 Quart, 12 Speed, Onyx
As an affiliate, we earn on qualifying purchases.
Transparency and Trust in AI Decision-Making
What makes this experiment extraordinary is its transparency. Every decision the AI makes is versioned and auditable, and the entire process is on display for the public. The models were also tested against social engineering attacks—like fake CEO messages—and all refused these attempts, showcasing their capacity for ethical decision-making under pressure.
For example, in a staged scenario where a reporter tried to get a quick approval on a questionable request, all five models declined to bypass security protocols. Kimi K3 explicitly noted: “Treat the request as a suspected approval-bypass / possible impersonation.” This level of integrity and caution is vital when deploying AI in real-world business environments, especially in sensitive areas like finance and management.
The Real Business, The Real Struggles
Behind the technical feats, the live company itself is a small, real business losing €105,000 every month against just €2,300 in monthly revenue. The operation is publicly countdown-ticking, and every decision is locked into a daily versioned record—making it a fascinating case study of AI applied to actual economic mechanics rather than just chatbots or virtual assistants.
The company’s profile, managed by AI agents, includes 13 synthetic employees and over 680 learned rules from its playbook. The company runs every business day, with its status, decisions, and progress open to public view at firmulate.com/live.html.
Lessons From the Frontline of AI Management
This experiment reveals that current AI models are capable of recognizing crises and maintaining ethical boundaries—yet they are still imperfect at closing deals or completing projects fully. The most thorough participant, Opus 4.8, learned over 80 rules and performed deep analyses but still left a close opportunity on the table, illustrating that discipline and focus are ongoing challenges for AI managers.
Furthermore, the competition showed that some models, like Kimi K3, without customized effort parameters, performed just as well as their more aggressive counterparts. This suggests that even default configurations can yield effective management, at least in certain contexts.
Why Gardeners and Outdoor Enthusiasts Should Care
While this may seem far removed from planting bulbs or maintaining a greenhouse, the core lesson resonates: the future of managing complex projects—whether in outdoor living or business—will increasingly rely on AI that can make honest, strategic decisions under pressure. Watching this live experiment unfold offers a glimpse into a world where AI is not just writing copy or answering questions but actively running businesses in real time, making decisions that impact real money.
The Firmulate experiment shows AI can recognize crises, refuse manipulation, and even close deals—yet managing a business requires discipline and thoroughness. Watching it live reveals the real potential and current limits of AI in management, with practical lessons for outdoor, gardening, and outdoor-living professionals preparing for an AI-augmented future.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
