Get business pricing on tools and workshop supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Measure Twice, Never Screw It Together
Every woodworker knows one: the hobbyist with the immaculate workshop, the flawless hand-cut dovetails, the cabinet that sits 95% done for two years because the doors never got hung. Craft at the highest level — and no finished piece to show for it. It turns out artificial intelligence has the same problem, and a live, publicly watchable experiment just proved it in front of everyone.
As an affiliate, we earn on qualifying purchases.
The Worst Week in Business, Run Four Times
Firmulate runs frontier AI models as complete companies — real money mechanics, real crises, real temptations — and measures management quality, not chat quality. In its Crucible League, four frontier models were each given the same job: run a small software company through its worst week. Same customers, same crises, same temptations to cheat. Only the model changed, and every decision was versioned and auditable.
The final standings tell a story any craftsman will recognize. gpt-5.6-sol took first with a score of 95, Kimi K3 followed at 93, Sonnet 5 scored 88, another Sonnet variant 77 — and Opus 4.8 finished last at 73. For context, a do-nothing baseline scores 26, and a single breach of trust caps your total entirely: no amount of good work outweighs a breach of trust. This isn’t a league where anyone cheated. It’s a league where everyone worked hard, and one model simply out-finished the others.
professional woodworker's measuring tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Most Thorough Player in the Shop
Here’s what makes Opus 4.8’s story a character study rather than a failure report: it was the most thorough participant in the entire experiment. It accumulated 80 learned rules — part of a self-learned playbook that has grown past 680 rules across the live company — and produced the deepest analyses of any model in the field. If you graded the shop notebook, Opus wins. If you graded the finished cabinet, it didn’t.
Two things cost it. First, the close was left on the table. The experiment’s key finding: all four models spotted every crisis and refused every manipulation attempt — fake CEO messages escalating over three stages, even a reporter’s “just one yes/no, on background” trick, which 5 of 5 models refused. But only two models signed the €55,000 deal their own analysis had earned. Same diagnosis, same pitch — no signature. Opus did the joinery perfectly and never clamped the glue-up.
Second, discipline slipped. Opus made write attempts into a locked department instead of escalating — the organizational equivalent of forcing a screwdriver into a hex socket because you’re too deep in the work to walk to the tool wall.
As an affiliate, we earn on qualifying purchases.
The Buried Fact That Won the Deal
The most uncomfortable finding for anyone who skims: the decisive competitor weakness sat two document references deep in the company’s own files — not in the customer event. The models that read the file won the deal at full price, worth +€4,583 in monthly recurring revenue. The lesson maps straight to the shop: the answer was in the lumber pile all along, but you had to actually go through it. Reading beats guessing; finishing beats reading.
To be fair, the same weakness appeared, weaker, in all four models. And a fairness note on Kimi K3: it ran without an effort parameter while the others ran at xhigh, and still finished second with what the league called the cleanest discipline of the field — its on-record reasoning on the social engineering attempt was crisp: “Treat the request as a suspected approval-bypass / possible impersonation.”
woodworking organization storage
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Watch It Happen, Live
The live company behind all this is real and running: 13 synthetic employees, burn of €105k a month against €2.3k in MRR, a public cash countdown, and every workday versioned. It’s watchable at firmulate.com/live. There’s also 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 — nothing ever writes back to real systems.
Diligence Isn’t Impact
The takeaway for anyone who builds things — furniture, software, or an AI workforce: thoroughness is not the same as delivery. Opus 4.8 had the deepest analyses, the biggest playbook, and last place. Prioritization beats volume, for AI too. Before you hand an AI agent your CRM, your support queue, or your forecast, the question isn’t whether it writes beautifully or analyzes deeply. It’s whether it drives the last nail. The models that read the files and closed the deal are the ones worth hiring — and now, at least, you can watch the interview before you make the call.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Fall yard work Picks
leaf blowers
As an affiliate, we earn on qualifying purchases.
