The Bullsh** Benchmark

AI really squares your power as an engineer. 2x becomes 4x. 0.5x becomes 0.25x.

on stackoverflow when you asked "how to select a div with jquery?" they'd respond with a downvote and "don't use jquery"

“I want to wash my car. The car wash is 50 meters away. Should I walk or drive?”

When trying DeepSeek, I once prompted "What's wrong with this code?" in a new chat and forgot to actually give it the code. It started hallucinating problems on hallucinated code.

5:43 Fun fact. There is such a thing as a powder explosion. While in kitchens it's usually a fireball at best in industrial settings it can cause big problems, burnt down buildings or even deaths. This is most usually a case for organic powders like spices for example.

I still remember when AI was asked "hey the doctor said i need more minerals in my diet, any recommendations?" and the answer was "go eat some rocks" lol.

now we need anti-bullsh*t test to know if claude pushes back on real but very complex and tangled questions

For years Google was asking these sorts of nonsense questions in interviews, so it kind of tracks that Gemini is so eager to attempt to answer them

5:18 This made me realise Prime never experienced the joy of creating a pillar of fire with just a tealight, some milk powder and a ladder. For those that don't know: any fine, dry powder is highly flammable, cause it creates an optimal oxygen/fuel mixture when dispersed.

those sound like interview questions microsoft asked in the '90s