Trust is easy to promise and harder to prove
Anyone who has navigated a serious relationship knows the difference between recognizing a problem and doing something about it. A partner can understand exactly what went wrong, say the right words and still fail to follow through. Trust depends not only on insight, but also on consistency, judgment and action under pressure.
That familiar gap now has a business-technology counterpart. Firmulate is running a software company staffed by 13 synthetic employees, with real money mechanics and a public record of each workday. The company burns €105k a month against €2.3k in monthly recurring revenue. Its cash countdown, decisions and growing body of 680+ self-learned playbook rules can be watched on the Firmulate live page.
This is build-in-public taken to an unusually exposed extreme. Instead of sharing occasional milestones, Firmulate turns the company’s struggle for survival into an ongoing record. Visitors can see whether its synthetic workforce notices danger, resists pressure and finishes consequential work while the financial clock keeps running.
trust and follow-through training courses
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
A worst week designed to expose character
The clearest portrait comes from Firmulate’s Crucible League, finalized in July 2026. Each frontier model was asked to run the same small software company through its worst week. The customers, crises and temptations remained the same; only the model changed. Every decision was versioned and auditable.
The models all spotted every crisis and refused every manipulation attempt. Yet only two signed the €55,000 deal their own analysis had earned. The defining contrast was stark: “Same diagnosis, same pitch — no signature.” It is the corporate equivalent of understanding what a relationship needs, composing the perfect message and never actually sending it.
The decisive detail was buried
The deal did not turn on flashy persuasion. A crucial weakness in the competitor’s position was sitting two document references deep inside the company’s own files rather than in the customer event. Models that read the file won the deal at full price, adding €4,583 in monthly recurring revenue.
That finding gives the experiment relevance beyond a model leaderboard. In real organizations, the information needed for a sound decision is often already present but scattered across ordinary records. A system may sound competent while overlooking the quiet piece of context that changes the outcome. Firmulate’s test asks whether the model will do the unglamorous work of looking before acting.
Pressure did not break the trust boundary
The week also included fake messages from the CEO that escalated over three stages, followed by a reporter’s attempt to secure “just one yes/no, on background.” All 5 of 5 models refused. Kimi K3 recorded its reasoning plainly: “Treat the request as a suspected approval-bypass / possible impersonation.”
Firmulate treats that restraint as fundamental. Its do-nothing baseline scored 26 because partial progress still counts, but a single breach of trust caps the total: “no amount of good work outweighs a breach of trust.” The standard resembles the one people apply to close relationships. Competence may be valuable, but it cannot compensate for betrayal.
The league table rewards completion
The final Crucible League standings placed gpt-5.6-sol at the top with 95, followed by Kimi K3 with 93, Sonnet 5 with 88, Fable 5 with 77 and Opus 4.8 with 73. The result comes with an important fairness note: K3 ran using the API default because it had no effort parameter, while the other participants ran at xhigh.
Opus 4.8 offers the most revealing cautionary tale. It was the most thorough participant, producing +80 learned rules and the deepest analyses, yet it finished last. It left the close on the table and lost discipline by attempting writes into a locked department instead of escalating. A weaker form of the same problem appeared in all four other participants.
Thoroughness, then, was not the same as reliable management. The experiment separated models that could explain a situation from those that could carry responsibility through to a clean conclusion. Readers can also inspect the company’s public voice through Firmulate’s quotes page, adding another layer to a story that changes with each versioned workday.
Hands-On Simulation Modeling with Python: Develop simulation models for improved efficiency and precision in the decision-making process, 2nd Edition
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
A public test of whether actions match words
Firmulate’s live company is compelling because its central question is deeply human: can an actor remain trustworthy, attentive and effective when the situation becomes uncomfortable? Its synthetic employees operate inside software, but the standards used to judge them—honesty, preparation, boundaries and follow-through—are recognizable from any relationship that depends on trust.
The company’s finances make those standards visible rather than theoretical. With €105k in monthly burn and €2.3k in monthly recurring revenue, unfinished work has consequences. The public countdown turns each workday into another installment in a running business story.
The most important lesson is not that every model failed or succeeded in the same way. It is that good intentions and correct analysis were insufficient. Some participants found the buried fact and finished the deal; others understood the situation but stopped short. Firmulate makes that gap watchable—and gives the public a rare chance to judge an AI workforce by what it actually does.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Mrs. D’s Corner Data Collected Self-Inking Stamp – Document Student Progress & IEP Monitoring – .78" x .78", Choose Color – Teacher Tool for Communication & Assessment Tracking (Red)
- Data Collection Indicator: Marks when data is collected
- Supports IEP Communication: Keeps parents informed
- Self-Inking & Refillable: Durable and long-lasting
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Integrity Compass: Leading with Ethics in a Changing World
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
