I'm working on a canvas that records your pen/brush strokes as you draw on it. https://superpaper.netlify.app/

Motivation is that large-language models have a very straight-forward task of predicting the next token and the dataset is easy to get. With this app I aim to do two things: 1. Gather a fairly large dataset that captures the brush-strokes for various art prompts. 2. Bootstrap an algorithm / model that can decompose any image/art/illustration into brush strokes.

A longer-term goal for this app is to build an auto-complete (Co-pilot or Grammarly equivalent) for art.

ps: This app has some bugs. Keep low expectations

FAANG is to be blamed for rapid rise in incompetence in tech. So many engineers are grinding on leetcode and bagging jobs that pay them insanely high salaries. These engineers know next to nothing when it comes to real world engineering challenges. I knew a guy who had a 200 day leetcode streak but didn't know the difference between a process and thread. We have essentially collectively decided that these stupid hiring processes are an accurate proxy for finding real talented people. I know so many colleagues who are here just to virtue signal that they work for a "cool" company. And it has become super easy to get into these companies and coast.

This is great but I'd like to see an autopilot that can do the hardest crash

Don't do image hashing please. Do this:

1. Get CLIP embeddings for text & images 2. Put them in a vector database (Pinecone.io or something similar)

It's unreasonably effective. Checkout this search engine: https://same.energy/