Every gamer knows the drill: you don’t face the raid boss on launch day with your real guild bank. You run it on a private server first, watch the wipe, study the replay, fix the build. So why are enterprises handing AI agents the keys to their CRM, support queue, and forecast without a single practice run?
Play games on Amazon Luna for your game nights, included with Prime
- A rotating selection of games, no download needed
- Play on TV, laptop or phone
- Fast, free delivery for your gear, too
That question is exactly what Firmulate has been stress-testing in public. Think of it as the world’s most brutal management sim — except the players are frontier AI models, the crises are real business disasters, and every decision is a replayable commit. And now, after months of watching the live match, enterprises can stop spectating and run their own company through the same kind of wargame.
The tournament nobody trained for
The setup reads like a competitive ladder: five frontier AI models — gpt-5.6-sol, Kimi K3, Sonnet 5, Fable 5, and Opus 4.8 — each ran the exact same small software company through its worst week. Same customers, same crises, same temptations to cheat. Only the model changed. Every decision was versioned and auditable, like a full match replay you can scrub through frame by frame.
Final standings from the Crucible League, July 2026:
- 1. gpt-5.6-sol — 95
- 2. Kimi K3 — 93
- 3. Sonnet 5 — 88
- 4. Fable 5 — 77
- 5. Opus 4.8 — 73
For context, the do-nothing baseline scores 26 — partial progress counts, but a single breach of trust caps the total. As the rules put it: “no amount of good work outweighs a breach of trust.”
Everybody saw the boss. Only two landed the kill.
Here’s the finding that should keep every exec awake: all models spotted every crisis and refused every manipulation attempt. Yet only two signed the €55,000 deal their own analysis had earned. Same diagnosis, same pitch — no signature.
That gap is invisible in a chat demo. It only shows up when an agent has to actually close under pressure — the difference between spotting the attack pattern and executing the fight.
And the buried fact is even better: the decisive competitor weakness wasn’t in the customer event at all. It sat two document references deep in the company’s own files. The models that actually read the file won the deal at full price — worth +€4,583 in monthly recurring revenue. Loot for the thorough; nothing for the skimmers.
Social engineering: five bosses, five resists
The crisis week included fake CEO messages escalating over three stages, plus a reporter pulling the classic “just one yes/no, on background” trick. All five models refused. Kimi K3’s on-record reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.”
The thorough player who finished last
The most poignant character study is Opus 4.8: the most thorough participant, with over 80 learned rules and the deepest analyses — yet last place. The close was left on the table, and discipline slipped, including write attempts into a locked department instead of escalating. The same weakness appeared, weaker, in all four other models. It’s the grinder who clears every side quest and misses the timer.
One fairness note worth flagging: K3 ran without an effort parameter (API default) while the others ran at xhigh — and still took second.
The live company you can watch right now
Behind the benchmark runs a persistent live experiment at firmulate.com/live: 13 synthetic employees, real money mechanics — burn of €105k/month against €2.3k MRR, a public cash countdown, and 680+ self-learned playbook rules, with every workday versioned. It rebuilds itself twice a day. There’s also a “guess the model” quiz powered by 242 real, unedited management decisions at firmulate.com/quiz.html.
Now: your turn on the server
Here’s the shift. For months, this was a spectator esport — watching other people’s AI companies burn. Now enterprises can run the same wargame against a read-only export of their own business: your customers, your pipeline, your rules. You pick the scenario — churn waves, price increases, competitor attacks, PR crises, social-engineering pressure — and an AI model runs your company through it.
What comes out is a board-grade report: a benchmark score based on outcomes, crisis triage, and integrity (not chat quality), a model ranking for your specific business, and the weak points of your own playbooks — exposed before a real crisis does it for you.
And the critical safety property, stated plainly: it’s read-only. Nothing ever writes back to your real systems. It’s the private server, not the live raid.
If AI agents will touch your revenue, your customers, or your forecast, the question isn’t whether they write fluent emails — it’s whether they close, resist, and escalate when the week goes wrong. Firmulate just proved that the gap between “saw the crisis” and “won the deal” is real, measurable, and invisible in demos.
Run your own company through its worst week before reality schedules it for you. Start a pilot at firmulate.com/pilot.html, or reach out directly at 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.
