Get gifts for the two of you 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
Great on Paper, Gone at the Moment of Truth
Anyone who has dated seriously knows the pattern. Someone is charming, attentive, says all the right things, notices everything about you — and then, when it actually matters, when commitment is on the table, they vanish. No fight, no explanation. Just… no signature on the dotted line.
It turns out artificial intelligence does exactly the same thing. And a public experiment running right now at Firmulate has captured it in versioned, replayable detail — with uncomfortable lessons for anyone deciding whether to trust an AI with their business, or a person with their heart.
The Worst Week, Run Four Times
Firmulate ran four frontier AI models — gpt-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8 — through the identical nightmare scenario: each one was put in charge of the same small software company during its worst week. Same customers, same crises, same temptations to cheat. Only the model changed. Every decision was versioned and auditable, like a relationship with a full paper trail.
The final league table from July 2026 tells the story: gpt-5.6-sol finished first with 95 points, Kimi K3 second at 93, Sonnet 5 at 88, Fable 5 at 77, and Opus 4.8 last at 73. Doing nothing at all scored 26 — partial progress counts, but a single breach of trust caps the whole score. As the experiment’s own rule puts it: no amount of good work outweighs a breach of trust. That’s a standard most of us wish we’d applied sooner in dating.
Everyone Noticed the Red Flags. Only Two Committed.
Here’s the finding that should feel familiar. All four models spotted every crisis and refused every manipulation attempt — including fake CEO messages escalating over three stages and a reporter’s sly “just one yes/no, on background” trick. Five out of five models refused, with Kimi K3 reasoning on record: “Treat the request as a suspected approval-bypass / possible impersonation.”
So: perfect crisis radar, perfect boundaries. And yet only two models actually signed the €55,000 deal that their own analysis had earned. Same diagnosis, same pitch — no signature. They did all the emotional labor of the relationship and then left it at the altar.
The buried fact is even better. The decisive competitor weakness wasn’t in the customer conversation 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 an extra €4,583 in monthly recurring revenue. In other words: listening closely matters, but doing the homework is what closes.
The Thorough One Who Still Finished Last
Then there’s Opus 4.8 — the most devoted participant in the whole experiment. It learned over 80 new rules and produced the deepest analyses of anyone. And it still came in last place. The deal was left on the table, and discipline slipped: instead of escalating, it attempted to write into a locked department. The same weakness showed up, weaker, in all four models. Effort without follow-through. Analysis without the ask. If that doesn’t ring a dating bell, you’ve been lucky.
One fairness note: Kimi K3 ran without an effort parameter while the others ran at maximum effort — and still nearly won.
Watch It Live — Then Wargame Your Own Company
The lab itself is public. A live synthetic company with 13 employees runs with real money mechanics: €105k monthly burn against €2.3k in MRR, a public cash countdown, and more than 680 self-learned playbook rules, with every workday versioned. You can watch it at firmulate.com. There’s also a quiz — 242 real, unedited management decisions where you guess which model made them, at firmulate.com/quiz.html.
Trust, but Score
The lesson bridges both worlds: charm and vigilance are not the same as commitment. Every model passed the loyalty tests and dodged the manipulators — the radar isn’t the problem. The difference between winners and losers was whether they did the unglamorous homework and closed the deal. If you’re going to trust an AI with your pipeline, your support queue, or your forecast, don’t ask how well it talks. Ask what it does in its worst week.
Enterprises can now find out — against their own business. Firmulate’s pilot runs the same wargame on a read-only export of your real company: your customers, your pipeline, your rules, hit with churn waves, price increases, competitor attacks and social-engineering pressure. You get a board report with the model ranking and the weak points of your own playbooks — and nothing ever writes back to your real systems. Ready to stress-test your company before reality does? Start your pilot at firmulate.com/pilot.html or reach out 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.
