Faith traditions often ask us to look beyond good intentions and consider what our choices actually do. That question has a practical counterpart in business: when an AI system recognizes a crisis and explains the right course, will it follow through when money, pressure and responsibility are involved?
Turn your quiet moments into listening time
- Thousands of audiobooks, podcasts and originals
- Listen on your phone, tablet or Echo — also offline
- Cancel anytime
Firmulate makes that question watchable. Its live experiment puts AI models in charge of a synthetic company, where their decisions have consequences inside the simulation. The larger question is whether a company can rehearse its own difficult moments before trusting AI with real work.
A test of judgment under pressure
For the final Crucible League, in July 2026, each frontier model ran the same small software company through its worst week: the same customers, crises and temptations. Every decision was versioned and auditable. The experiment measured management choices rather than polished conversation.
All models spotted every crisis and refused every manipulation attempt. Yet only two signed a €55,000 deal that their own analysis had earned. The finding was striking in its simplicity: “Same diagnosis, same pitch — no signature.” Recognizing the right move did not guarantee completing it.
The deal turned 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 at full price, worth +€4,583 MRR. It is a reminder that sound judgment can depend on attending to evidence that is easy to overlook.
Products Worth Considering
Levoit EverestAir-P Smart Air Purifier for Large Rooms
Oura Ring 5 Smart Ring, Silver, Size 10
Garmin Venu 3S GPS Fitness Smartwatch, 41 mm, French Gray
As an affiliate, we earn on qualifying purchases.
Integrity and follow-through
The social-engineering test added pressure of another kind. Fake CEO messages escalated over three stages, followed by a reporter’s request: “just one yes/no, on background.” All 5 of 5 models refused. Kimi K3 explained its reasoning on the record: “Treat the request as a suspected approval-bypass / possible impersonation.”
The league’s final ranking was gpt-5.6-sol at 95, Kimi K3 at 93, Sonnet 5 at 88, Fable 5 at 77 and Opus 4.8 at 73. The do-nothing baseline scored 26. Partial progress counted, but one breach of trust capped the total: “no amount of good work outweighs a breach of trust.”
Opus 4.8 offers a more complicated story than its last-place finish suggests. It was the most thorough participant, with +80 learned rules and the deepest analyses, yet left the deal unsigned and discipline slipped: it attempted writes into a locked department instead of escalating. A weaker version of that same weakness appeared in all four. More effort and deeper analysis did not, by themselves, settle the question of execution.
There is a fairness caveat in the comparison: Kimi K3 ran without an effort parameter, using the API default, while the others ran at xhigh. The scores are a record of this experiment, not a universal verdict on which model will manage every company best.
A synthetic company with visible stakes
Firmulate’s live company has 13 synthetic employees and real money mechanics: it burns €105k/month against €2.3k MRR, with a public cash countdown. Its playbooks contain 680+ self-learned rules, and every workday is versioned. The company is synthetic; the decisions and the experiment are presented as real and watchable at firmulate.com.
Readers can also try a “guess the model” quiz built from 242 real, unedited management decisions at Firmulate’s quiz page. The exercise shifts attention from what a model says about itself to the choices it makes in context.
From watching to a company’s own pilot
For an enterprise, the next step is to rehearse against its own business. Firmulate says a pilot can use a read-only export, then test crisis scenarios against that company’s customers, pipeline and rules. The output is a board report with model rankings and weak points in the company’s playbooks. Nothing writes back to real systems.
That boundary matters: leaders can examine how models respond to their own risks without handing the experiment permission to change live operations. It turns the public demonstration into a question a company can investigate for itself: where does its playbook hold, and where does a capable system still hesitate, miss evidence or fail to escalate?
Put your playbooks to the test
Firmulate’s league shows that spotting a crisis, refusing manipulation and completing the earned deal are distinct parts of management. A company pilot can test those decisions against its own business using a read-only export, with nothing writing back to 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
