Counteraxiom is founded on the basis that much of the AI on the market never serves as constructive criticism, and was too “shy” to warn users against poor decisions — even going as far as to further advocate for the motion.
You'd paste a half-baked argument into any mainstream chatbot and it would come back more eloquently restructured. Sometimes humans require an unbiased source to accurately describe and assess their situation, or to validate their ideas.
The frontier labs are racing to make their models more helpful, more agreeable, more eager-to-please. That's great for productivity. It's terrible for thinking.
The smarter the tools got, the more validated people felt. Ever had friends pitching you stuff that any decent counter-argument would shred, sounding more sure of themselves than they'd ever been before they started using AI?
Counteraxiom isn't supposed to be your assistant. It's supposed to be the friend who tells you your idea is dumb before you spend six months on it. I use it every day. The day I stop wanting to argue with Counteraxiom is the day the product is broken.
— Kyle Hui