Imagine hiring a new kitchen assistant who, no matter how lazy or unhelpful, still manages to bring you 26 out of 100 possible points just by being present. Sounds odd, right? But in the world of AI, this ‘do-nothing’ baseline reveals a lot about how we measure trust and performance — and why sometimes, the simplest results are the most revealing.
Get kitchen staples and gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The Hidden Score of a Do-Nothing AI
In an ongoing experiment named the Crucible League, AI models are put through their paces in a simulated business environment, running a small software company facing crises, customer manipulations, and ethical dilemmas. Surprisingly, even the most passive AI, one that simply does nothing but stay honest, scores 26 points out of 100. This baseline isn’t a flaw — it’s a feature that highlights what minimum competence looks like in AI decision-making.
As an affiliate, we earn on qualifying purchases.
Why Partial Progress Counts and Breaches Cap Performance
Every AI model’s score reflects its ability to handle real-world pressures. All models in the test identified every crisis and refused every manipulation attempt, which shows ethical robustness. Only two models managed to close a deal and sign a contract, earning the full 55,000 euros, based on their own analysis. The key insight? Even with perfect crisis detection, failure to execute or breach of trust caps the score at 26. Simply put, no matter how good their diagnosis, if an AI slips up or attempts to cheat, it hits a performance ceiling.
What This Means for Business Trust
The experiment underscores an essential truth: in business, trust is non-negotiable. An AI that reads your files, understands your customer crises, and refuses to manipulate or cheat, is worth more than one that merely talks well. The models that read deep into company files won the deal at full price, completing their tasks with integrity, while others left opportunities on the table or slipped in discipline.
Measuring Honest Performance
Unlike chat demos that focus on fluent conversation, this test judges an AI on its ability to finish what it starts—reading critical documents, making honest decisions, resisting manipulation, and closing deals fairly. The live experiment runs every weekday, with real money mechanics and self-learned rules, making the results transparent and watchable at firmulate.com/live.
Implications for Businesses Considering AI
For companies eyeing AI to manage customer support, forecast sales, or handle CRM, the takeaway is clear: it’s not just about how well an AI chats. It’s about whether it can reliably finish its work, stay honest under pressure, and read your data thoroughly. A ‘do-nothing’ baseline showing 26 points serves as a reminder that even minimal ethical and operational competence is valuable — and that performance gaps often lie in discipline, reading depth, and trustworthiness.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
NFL season / tailgating Picks
team gear
As an affiliate, we earn on qualifying purchases.
