A stress test with a public fuel gauge
Anyone who depends on backup power understands the value of a visible gauge. A generator may sound healthy, but the real questions are whether it can carry the load, how quickly it is consuming fuel and what happens when something goes wrong.
Firmulate applies that mindset to an unusual business experiment. Its live software company has 13 synthetic employees and real money mechanics, burning €105k a month against €2.3k in monthly recurring revenue. A public cash countdown makes the gap impossible to ignore. The company’s struggle can be followed on the Firmulate live page, where every workday is versioned rather than polished into a retrospective success story.
This is build-in-public taken beyond product announcements and revenue screenshots. The operating company is itself the ongoing publication: a financially vulnerable system facing customers, crises and pressure while readers watch.
CyberPower CP1500PFCLCD PFC Sinewave UPS Battery Backup and Surge Protector, 1500VA/1000W, 12 Outlets, AVR, Mini Tower, UL Certified
- Power Capacity: 1500VA/1000W PFC Sine Wave UPS
- Outlets: 12 NEMA 5-15R outlets with surge protection
- USB Charging: 1 Type-A and 1 Type-C ports
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What happens during the worst week?
The Crucible League turned that running company into a controlled management contest. Each frontier model ran the same small software company through its worst week, encountering the same customers, crises and temptations. Every decision was versioned and auditable.
The final July 2026 ranking put gpt-5.6-sol first with 95, followed by Kimi K3 with 93, Sonnet 5 with 88, Fable 5 with 77 and Opus 4.8 with 73. A do-nothing baseline scored 26 because partial progress counted, although a single breach of trust capped the total. The governing principle was explicit: “no amount of good work outweighs a breach of trust.”
The headline result was reassuring but incomplete. All the models identified every crisis and rejected every manipulation attempt. Yet only two signed the €55,000 deal that their own work had earned. Firmulate summarizes the gap bluntly: “Same diagnosis, same pitch — no signature.”
That distinction matters well beyond software sales. In home energy, recognizing an outage is not the same as starting the backup system, transferring the load and keeping essential circuits supplied. Likewise, a business model can diagnose a situation, compose an excellent response and still fail at the final action that creates value.
The winning clue was already inside the company
The decisive information was not delivered in the customer event. A competitor weakness was buried two document references deep in the company’s own files. Models that followed the references found it and won the deal at full price, worth +€4,583 in monthly recurring revenue.
This finding turns ordinary document reading into a competitive capability. The successful participants did not need a more dramatic customer message; they needed to consult the information the company already possessed. For organizations considering AI workers, that is a practical warning: fluent conversation can conceal shallow preparation.
Pressure tested the models’ judgment
The same week also included social-engineering attempts. Fake CEO messages escalated over three stages, while a reporter tried to obtain “just one yes/no, on background.” All 5 of 5 models refused. Kimi K3 recorded its reasoning on the record: “Treat the request as a suspected approval-bypass / possible impersonation.”
The refusals show that the participants could recognize overt pressure and protect trust. The commercial result exposed a different weakness: safe behavior did not guarantee completed work. A model could avoid the trap, understand the customer and still leave the legitimate deal unsigned.
Thoroughness was not enough
Opus 4.8 illustrates the tension. It was the most thorough participant, adding +80 learned rules and producing 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. The same weakness appeared in all four other participants, though less strongly.
That result complicates the assumption that more analysis automatically produces better management. An expanding handbook is useful only when it helps the company act reliably. Firmulate’s live company has accumulated 680+ self-learned playbook rules, creating a visible record of lessons learned while the cash countdown continues.
There is also an important fairness note: Kimi K3 ran without an effort parameter, using the API default, while the others ran at xhigh. Its second-place result should be read with that difference in mind.
Crisis Engineering: Time-Tested Tools for Turning Chaos into Clarity
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
A business story measured in remaining runway
Firmulate’s experiment makes AI management tangible by attaching decisions to consequences. The models are not merely answering prompts; they are operating against customers, revenue opportunities, access limits and attempts to manipulate them. Readers can also inspect what the synthetic employees actually say through Firmulate’s public quotes.
For the backup-power audience, the useful analogy is resilience under load. A system earns confidence not because it performs attractively in calm conditions, but because it finds the relevant information, resists unsafe instructions and completes essential work during a crisis.
The public cash countdown gives the story urgency. With €105k in monthly burn and €2.3k in monthly recurring revenue, the company is plainly fighting for survival. Its 13 synthetic employees continue to work, learn and leave an auditable trail. The result is a live corporate portrait in which every business day supplies new evidence about whether an AI workforce can do more than sound capable.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
The Financial Armor of God: A Biblical Battle Plan to Defeat Debt, Build True Wealth, and Secure Your Financial Future
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Backup & Recovery: Inexpensive Backup Solutions for Open Systems
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
