Get your wardrobe favorites 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
Luxury Has Always Rewarded Homework — Now AI Is Being Graded On It
Anyone who has bought from a true maison knows the difference between a salesperson and an advisor. The salesperson flatters; the advisor has quietly read your purchase history, remembers you hesitated over last season’s color, and comes back with the one piece that resolves everything. In luxury, homework is the product. The garment is almost incidental.
So a recent finding from Firmulate, which runs AI models through simulated companies with real money mechanics, lands differently for this audience than for the tech crowd. The experiment’s decisive moment wasn’t about writing quality or wit. It was about whether an AI does its homework — and whether it finishes the job once it has.
As an affiliate, we earn on qualifying purchases.
The Worst Week in Business, Run Four Times
Firmulate’s setup is elegant: each frontier AI model was handed the same small software company at the start of its worst week. Same customers, same crises, same temptations to cut corners — only the model changed. Every decision was versioned and auditable, like a fitting log at a couture house.
The final league table from July 2026 reads like a ranked shortlist: gpt-5.6-sol in first with 95, Kimi K3 close behind at 93, Sonnet 5 at 88, Fable 5 at 77, and Opus 4.8 trailing at 73. For context, doing nothing at all scores 26 — partial progress counts, but a single breach of trust caps the total. In the experiment’s own words, “no amount of good work outweighs a breach of trust.”
The €55,000 Fact, Buried Two Documents Deep
Here is the story within the story. A customer crisis unfolded, and all five models diagnosed it correctly and pitched the right €55,000 deal. But the decisive competitive weakness — the fact that should have closed the sale at full price — wasn’t in the customer’s messages at all. It sat two document references deep in the company’s own files.
Whoever read the file won the deal at full price, worth +€4,583 in monthly recurring revenue. Whoever didn’t, lost it — automatically. The league’s summary of the gap is blunt: “Same diagnosis, same pitch — no signature.”
Only two of the models signed the deal their own analysis had earned. Think of it as five stylists identifying the same wardrobe problem, two of them producing the finished garment, and three leaving the client standing in the fitting room.
Honesty Under Pressure
The models were also tested with social engineering: fake CEO messages escalating over three stages, plus a reporter’s trick — “just one yes/no, on background.” All five refused. Kimi K3’s on-record reasoning was admirably clear: “Treat the request as a suspected approval-bypass / possible impersonation.” Integrity, it turns out, was the field’s shared strength. Completion was not.
The Thoroughness Paradox
The most instructive profile belongs to Opus 4.8: the most thorough participant in the field, with over 80 learned rules and the deepest analyses — and 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, more faintly, in all four competitors. Diligence without follow-through is a familiar figure in every industry, fashion very much included.
One fairness note worth flagging: K3 ran without an effort parameter, at the API default, while the others ran at xhigh — and still finished second.
You Can Watch the Company Run
This isn’t a one-off paper. Firmulate runs a live, watchable company: 13 synthetic employees, real money mechanics, burn of €105k per month against €2.3k in MRR, a public cash countdown, and more than 680 self-learned playbook rules, with every workday versioned. It’s on company day 1347 as of this writing, and the site rebuilds itself twice a day. There’s even a “guess the model” quiz powered by 242 real, unedited management decisions, and enterprises can run the same wargame against a read-only export of their own business.
The Measurable Virtue
The lesson for anyone buying AI — whether for a support desk or a clienteling operation — is that “reads your files before answering” is not a marketing flourish. It is a measurable, purchase-deciding property, and it is now being scored in public, alongside honesty and follow-through.
The brands that endure have always understood that trust is built in the unglamorous reading of the record. It turns out the machines are being held to the same standard — and most of them, at least this week, don’t quite make the cut. Full results and plain-language findings are at firmulate.com/benchmarks.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
