OpenAI research loop was a joke honestly

OpenAI research loop was a joke honestly

Just got rejected from OpenAI's research track after two rounds of ML coding, and I gotta vent because the bar is absurd.

First round was 1 NN with L2 distance, had to vectorize it in numpy with zero loops then reformulate the whole thing as a neural network forward pass. Fine, doable. Second round was entropy calculation with the log sum exp trick for numerical stability, then they flip it into a streaming version where logits show up block by block. I got both working.

Interviewer barely reacted the whole time. Felt like talking to a wall. Then a week later, rejected, no real feedback beyond "close call."

I missed my chance at generational wealth over what, a slightly slower streaming implementation? Give me a break. Everyone acts like this place has some sacred bar but it's really just vibes and whoever the interviewer felt like liking that day.

Anyone else get the ML coding rounds and bomb despite finishing everything? Curious if this is just me or if OpenAI research track is genuinely this arbitrary.