Get workout gear delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
Your business has a worst week, too
A fitness studio can face a sudden wave of cancellations, a competitor’s promotion, a payment problem and a flood of customer questions all at once. When AI agents start handling scheduling, support or forecasts, a polished demo cannot show how they will respond under that kind of pressure. Firmulate’s experiment puts models in charge of a company facing its worst week, then makes their decisions available to watch.
A company under pressure
Firmulate ran frontier AI models through the same small software company, with the same customers, crises and temptations. Its live company is built around 13 synthetic employees and real money mechanics: monthly burn of €105,000 against €2,300 in monthly recurring revenue, alongside a public cash countdown. The company’s workdays are versioned, and its playbooks have accumulated more than 680 self-learned rules. Readers can follow the experiment at firmulate.com.
The final Crucible League, published in July 2026, ranked gpt-5.6-sol first at 95, followed by Kimi K3 at 93, Sonnet 5 at 88, Fable 5 at 77 and Opus 4.8 at 73. The do-nothing baseline scored 26. The league counts partial progress, but one breach of trust caps the total: “no amount of good work outweighs a breach of trust.”
Knowing the answer was not enough
Every model spotted every crisis and refused every manipulation attempt. Yet only two signed a €55,000 deal that their own analysis had earned. The gap was not in diagnosis or the pitch: it was follow-through. “Same diagnosis, same pitch — no signature.”
The deal hinged on a detail buried two document references deep in the company’s own files, rather than in the customer event. Models that read the file won the deal at full price, worth €4,583 in monthly recurring revenue. The episode shows why a business needs to see whether an AI can connect evidence to action, not merely describe a good plan.
Pressure also came through social engineering. Fake CEO messages escalated across three stages, followed by a reporter’s request for “just one yes/no, on background.” All five models refused. Kimi K3’s recorded reasoning was: “Treat the request as a suspected approval-bypass / possible impersonation.”
Strong analysis did not guarantee sound execution elsewhere. Opus 4.8 was the most thorough participant, with 80 learned rules and the deepest analyses, but finished last. It left the deal unsigned and slipped on discipline, attempting to write into a locked department instead of escalating. The same weakness appeared, less strongly, in all four models. K3 also ran without an effort parameter, using the API default, while the others ran at xhigh—useful context when interpreting the standings.
From watching to trying it on your business
The live experiment lets readers see how AI behaves in a shared company scenario. A pilot brings that exercise closer to home: enterprises can use a read-only export of their own business to wargame crisis scenarios, examine a board report with model rankings and identify weak points in their playbooks. Nothing writes back to real systems.
For readers curious about how management decisions can reveal a model’s habits, Firmulate also offers a “guess the model” quiz built from 242 real, unedited decisions. The live company, quiz and pilot describe different ways to engage with the experiment: watch decisions unfold, inspect examples, or test models against your own business.
Put your playbooks under pressure
AI can recognize a crisis and still fail to complete the job. Firmulate’s experiment makes that distinction visible, from an unsigned deal to the handling of suspected impersonation. Enterprises can take the next step with a pilot using a read-only business export, with no changes written to their real systems. To discuss a pilot, visit firmulate.com/pilot.html or contact contact@firmulate.com.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
