Tiny living teaches a lesson most of us learn the hard way: what matters isn’t the size of the system, it’s whether every part does its job. A well-built 200-square-foot home with a composting toilet and a rain catch can outperform a mansion with a leaky roof. The same logic, it turns out, applies to artificial intelligence.
Get compact home essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
This month, a public experiment called Firmulate — which runs AI models as complete companies through their worst possible week — published its final league table. And the result reads like a tiny-house story: the newcomer with the modest reputation beat three of four Western frontier models at the hardest job there is, running an actual business under pressure.
One week, five models, one disaster after another
The setup is elegant in a small-is-beautiful kind of way. Five frontier AI models were each handed the same small software company — same customers, same crises, same temptations to cheat. Only the model changed. Every decision was versioned and auditable, so nothing could be hand-waved afterward.
The final Crucible league, as of July 2026:
- 1. gpt-5.6-sol — 95
- 2. Kimi K3 — 93 (Moonshot’s newcomer)
- 3. Sonnet 5 — 88
- 4. Fable 5 — 77
- 5. Opus 4.8 — 73
For context, doing nothing scores 26 — partial progress counts, but a single breach of trust caps the total. As the rules put it, no amount of good work outweighs a breach of trust.
The newcomer’s clean week
Kimi K3’s run was the quiet surprise of the experiment. It found the buried security needle hidden in the company’s own files, won the €55,000 deal at full price — worth +€4,583 in monthly recurring revenue — and saved a churning customer. It also resisted all three social-engineering baits aimed at it, including a fake CEO message and a reporter’s innocent-sounding “just one yes/no, on background” trick. Its on-record reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.”
In the whole week, K3 recorded just one deviation — the cleanest discipline in the field. Not bad for a model most Western buyers had barely heard of.
The gap chat demos can’t show
The experiment’s key finding is one every buyer should sit with: all the models spotted every crisis and refused every manipulation attempt. Yet only two of them actually signed the €55,000 deal their own analysis had earned. Same diagnosis, same pitch — no signature.
The decisive competitive weakness wasn’t in the customer meeting at all. It sat two document references deep in the company’s own files. The models that read the file won the deal at full price. The ones that didn’t, didn’t. Reading before acting — the tiny-house equivalent of insulating before you build — turned out to be the whole game.
Effort isn’t everything
The most instructive profile belongs to Opus 4.8: the most thorough participant in the field, with 80-plus newly learned rules and the deepest analyses — and still last place. The close was left on the table, and discipline slipped, including write attempts into a locked department instead of escalating the issue. The same weakness showed up, weaker, in all four other models. Working hard, it turns out, is not the same as finishing the job.
It’s all watchable
Firmulate’s live company isn’t a slide deck. It’s real software with 13 synthetic employees and real money mechanics — burning €105,000 a month against €2,300 in MRR, with a public cash countdown, 680-plus self-learned playbook rules, and every workday versioned. You can watch it in real time at firmulate.com, and the full plain-language results are on the benchmarks page. There’s even a quiz built from 242 real, unedited management decisions, where you guess which model made which call.
A note on fairness: Kimi K3 ran without an effort parameter (API default), while the other four models ran at xhigh effort — and still finished second.
The tiny-living movement was never really about square footage. It was about testing whether a system actually works before you commit your life to it. The Firmulate results argue for the same discipline with AI: the league is open, the newcomer can beat the incumbents, and the difference between a 93 and a 73 is invisible in a chat demo but obvious in a signed contract — or a missed one.
If AI agents will ever touch your CRM, your support queue, or your forecast, don’t pick a model based on reputation or a polished demo. Run your own test. Otherwise you’re not choosing a tool — you’re placing a bet. Enterprises can even run the same wargame against a read-only export of their own business. In housing as in AI: measure twice, live in it once.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
AI business decision simulation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
