Made this weapon configurator with Opus 5
Started with an image and asked Opus to recreate the gun using img2threejs, then turn it into an interactive customization experience
Google's gemma 4 scored 5 out of 5 on every single prompt he threw at it running locally, coding, math
Reasoning, all of it, while opus 5 bills $200/MO and GPT-5.6 bills $200/MO for the same answers 01:28 he scrolls his own eval grid and reads it off, "five out of five, five
This quietly ended the entire ai debate
Moon dev: "gpt 5.6 is the best at math and i can prove it in a way nobody expected" "i'm a die hard claude guy so watch me admit the thing i really did not want to admit" "opus 5 versus gpt 5.6, one versus two, and the gap is not what
Just imagine, $100K is on the line and Claude vs GPT have to decide where it goes.
Found Compelle instead of asking one AI a question and getting a confident answer, Compelle makes two AI agents take opposite sides and fight it out and the whole thing runs on Bittensor SN82
You have reached the end of the archive
All of Claude Opus 5