DeepSeek & ChaptGPT o1 & ChatGPT o3-mini coding test experiment results

Video Description

Go to https://iob.fyi/COCR to test yourself against the world's best AIs, while learning about the inner workings of core Internet technologies in the process. By popular demand, I've done some experimenting on how well (or not) the new DeepSeek-R1 AI can code, and how it compares to ChatGPT o1 and ChatGPT o3-many-high. This video has a summary of my results Stay tuned for an upcoming video where I discuss what I think the implications will be for DeepSeek on programmers for the next couple of years. 00:00 Two new AI models have dropped 00:49 Executive summary 02:05 Video Intro 02:34 o3-mini isn't as good as o1, and much slower 03:02 How the challenges work 04:16 Challenge 1: SQLite in Zig 06:21 Challenge 2: (Easier) SQLite in Go 09:25 Challenge 3: (much easier) DNS Server in Go 10:32 Wrap up If, for some unfathomable reason, you really want to watch the raw screen output (with no commentary) of these experiments, you can find them at: https://www.youtube.com/watch?v=9f9R4_XStIc (SQLite in Zig) https://www.youtube.com/watch?v=b6ZiOKMzwfA (SQLite in Go) and https://www.youtube.com/watch?v=lMHvrkx-6gk (DNS Server in Go) I'll have a video up for members eventually to discuss the code quality of these AIs.