Leetcode style problems do test raw IQ. Here's the thing though, you need to get really creative with EACH of your problems to dodge the people who just practic

Leetcode style problems do test raw IQ. Here's the thing though, you need to get really creative with EACH of your problems to dodge the people who just practice for these things as even an official IQ test is game-able with practice. If someone solves a problem that doesn't exist or isn't a permutation of an existing problem it is a sign of higher IQ. For example. Given a grid where each element is 1 or 0. Count the amount of donuts in the grid. A donut is a cluster of 1s with a cluster of 0s within it. Donuts can exist within donuts. Extend this problem to 3D to count the amount of hollow shells in a 3D grid. Given a grid where only a single cluster of 1s exist. Determine if that cluster forms a 3D donut. These problems I just created from scratch. A person who solves it in an interview, I know he is smart because I know he never looked it up. Here's the other thing. The skills involved in solving this problem is highly unrelated to the day to day tasks of a software engineer. Additionally I believe anyone can solve these problems (except for the last one) given enough time (and you certainly are given enough time on the job). The software interview is mainly testing for IQ based on how FAST you solve this problem. Either way, software engineering itself is easy, it doesn't need high IQ. But make no mistake. A person who solves these problems and does so both quickly and without pattern matching the problem to an existing one he/she practiced for is demonstrating higher intelligence which is what most interviewers at FAANG are basically trying (and arguably sort of failing) to test for.