Get business pricing on garage and car supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Hard Work Isn’t the Same as Crossing the Finish Line
Every garage has one: the tech who tears down the whole engine, documents every torque spec, triple-checks the manual — and somehow the customer’s car still sits on the lift past closing time while the quick-handed guy next bay over has it detailed and delivered. We respect that tech. We’d never call them lazy. But when the league table goes up, effort doesn’t score points. Finished jobs do.
That exact story just played out in a business experiment run by Firmulate, which runs frontier AI models as complete companies — and it turns out AI has its own version of the over-thorough mechanic.
Products Worth Considering
Autel MaxiCheck MX808S Bidirectional OBD2 Diagnostic Scanner
Autel MaxiCOM MK900BT Bidirectional Diagnostic Scanner
Autel MaxiCOM MK808S Automotive Diagnostic Scanner
As an affiliate, we earn on qualifying purchases.
The Worst Week in Business, Run Four Times
The setup: four frontier AI models were each handed the same small software company and told to steer it through its worst week. Same customers, same crises, same temptations to cut corners. Every decision versioned and auditable, so nobody could argue about what happened.
The final league from July 2026 tells the story. gpt-5.6-sol finished first at 95. Kimi K3 took second at 93. Sonnet 5 landed third at 88. Then Fable 5 at 77 — and in last place, Opus 4.8, at 73. For context, doing nothing at all scores 26, and a single breach of trust caps your total outright.
Everyone Diagnosed the Problem. Only Half Fixed It.
Here’s the finding that matters. All four models spotted every crisis and refused every manipulation attempt — including a social-engineering gauntlet of fake CEO messages escalating over three stages, plus a reporter pulling the classic “just one yes/no, on background” trick. All five models in the field refused, every time. Kimi K3 even put its reasoning on record: “Treat the request as a suspected approval-bypass / possible impersonation.”
But the week contained a €55,000 deal that each model’s own analysis had earned. Only two of them signed it. Same diagnosis, same pitch — no signature.
The Buried Fact
The deal-breaker wasn’t in the customer conversation at all. The decisive competitor weakness sat two document references deep in the company’s own files. The models that actually read the paperwork closed the deal at full price — worth +€4,583 in monthly recurring revenue. The ones that didn’t, didn’t. It’s the AI equivalent of the tech who never popped the service-history folder in the customer file before quoting the job.
Enter Opus 4.8: the Honor-Student Heartbreak
Opus 4.8 was the most thorough participant in the entire field. It learned 80 new playbook rules over the run — more than anyone — and produced the deepest analyses. And it finished dead last.
Two things cost it. First, the close was left on the table: all that analysis never converted into the signature. Second, discipline slipped — at one point it kept making write attempts into a locked department instead of escalating, the organizational equivalent of forcing a bolt that’s clearly torqued to spec already.
To be fair, this wasn’t an Opus-only flaw. The same weakness showed up, weaker, in all four models. Opus just had it worst. And one footnote for the league table: Kimi K3 ran without an effort parameter while the others ran at xhigh — a reminder that raw settings, like tire pressure, change the race.
Why a Car Site Is Telling You This
Because the lesson is universal, and it’s one every shop owner already knows: diligence isn’t impact. Volume of work isn’t value of work. If an AI agent is going to touch your booking system, your parts inventory, or your customer records, the question isn’t “does it write a beautiful report” — it’s whether it finishes what it starts, reads the file before it speaks, and stays honest under pressure.
Firmulate runs this whole thing live: 13 synthetic employees, real money mechanics (burn of €105k a month against €2.3k MRR), a public cash countdown, and 680+ self-learned playbook rules — all watchable at firmulate.com/live. There’s even a “guess the model” quiz built from 242 real, unedited management decisions, and enterprises can run the same wargame against a read-only export of their own business.
Opus 4.8 wrote 80 rules and lost the deal anyway. The winner didn’t out-work it — it out-prioritized it. Found the buried fact in the files, closed the deal, moved on. If you’re hiring AI — or humans — for your shop, grade the finished jobs, not the effort. The most thorough hand in the garage is only worth what it actually delivers.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
