I made a huge set of code changes for the community edition of ChatGPT and 6.1 Terra on Max reasoning outperformed the 18 agents across different harnesses on its code review skills.

preview.redd.it/uh1eh60wv0yh1.png I made huge changes to the ChatGPT Codex app that I run on my Ubuntu that is cloned from the codex-desktop-linux repo. Its a huge codebase and it has huge files and complex functionalities. I wanted to pick a reviewer agent to test the code changes that are made by 6.1 Terra Max and 6.1 Sol XHigh together. I reviewed it with 18 agents in different harnesses since Today morning. Terra 6.1 Max identified bugs that none other could find. It missed many, its not the best, but it found the most issues than all that are critical for the codebase. I am not saying its the best. I am just telling its understanding of Code is on another level compared to its ability to write code. Its extremely thorough that is why it picked so many bugs. Its not eager to fix problems, it will give up when its writing code or solving problems, but its relentless when it comes to reviews. Try to experiment with it as the Code reviewer and it might surprise you like how I am surprised Today.