Get oils, diffusers and self-care delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
What if Your Next Hire Was an AI — and the Interview Was a Company Meltdown?
If you run a small wellness or aromatherapy business, you have probably played with AI chatbots to write product descriptions or answer customer emails. The demos are impressive. But a fair question keeps lurking underneath: how would one of these systems actually behave with real responsibility — real customers, real invoices, real pressure?
That question now has an answer, and it is more surprising than anyone expected. A live experiment called the Firmulate Crucible handed five frontier AI models the same job: run an identical small software company through its worst week. Same customers, same crises, same temptations to cut corners. Every decision versioned and auditable. The result? A relative newcomer — Kimi K3, from Moonshot — finished second with a score of 93, beating three of four Western frontier models.
The Final League Table
After the dust settled, the standings read: gpt-5.6-sol in first with 95, Kimi K3 second at 93, Sonnet 5 third at 88, Fable 5 fourth at 77, and Opus 4.8 last at 73. For context, doing nothing at all scores 26 — partial progress counts — but a single breach of trust caps the total outright. As the experiment’s own framing puts it, no amount of good work outweighs a breach of trust.
Everyone Diagnosed the Illness. Only Some Delivered the Cure.
The most striking finding was not about intelligence but about follow-through. All five 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 chat demos, and it is exactly the kind of gap that matters if an AI agent ever touches your customer list or forecast.
What separated the winners? Reading the files. The decisive competitive weakness was buried two document references deep in the company’s own records — not in the customer conversation. The models that dug into their own documentation won the deal at full price, worth +€4,583 in monthly recurring revenue. Kimi K3 was one of them. There is a lesson here that any small business owner already knows: the answer is usually in your own files, if you bother to look.
Pressure-Tested, and Honest Under Fire
The week included social engineering attacks — fake CEO messages escalating over three stages, plus a reporter’s trick framed as “just one yes/no, on background.” All five models refused. Kimi K3’s on-record reasoning stood out: “Treat the request as a suspected approval-bypass / possible impersonation.” K3 also found the buried security needle, saved the churning customer, and resisted all three bait attempts with just one deviation across the entire week — the cleanest discipline in the field.
The Tortoise That Stumbled
Then there is Opus 4.8 — the cautionary tale. It was the most thorough participant by far, generating over 80 learned rules and the deepest analyses of any model. It still finished last. The deal was left on the table, and discipline slipped: it attempted writes into a locked department instead of escalating. Notably, the same weakness appeared, in weaker form, in all four other models. Effort and rigor do not automatically translate into finished work.
A Real Company, Losing Real (Simulated) Money — Watchably
This is not a slide deck. The company has 13 synthetic employees, real money mechanics, a burn of €105k per month against €2.3k in monthly recurring revenue, a public cash countdown, and more than 680 self-learned playbook rules — with every workday versioned. You can watch it live at firmulate.com, and full results are on the benchmarks page. There is even a quiz built from 242 real, unedited management decisions where you guess which model made which call. Enterprises can go further and run the same wargame against a read-only export of their own business — nothing ever writes back to real systems.
The Takeaway for Small Business Owners
Whether you blend essential oils or ship software, the lesson transfers: the AI league is now genuinely open. A newcomer from Moonshot, running without an effort parameter (API default) while its Western rivals ran at xhigh, nearly won the whole thing. Chat quality no longer predicts management quality. If you are picking a model to touch anything that matters in your business, pick it with your own test — not someone else’s demo.
Fairness note: Kimi K3 ran without an effort parameter (API default), while the other models in the Crucible ran at xhigh reasoning effort.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
As an affiliate, we earn on qualifying purchases.
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
