Very niche imo but: I am doing a Reverse Engineering Project for a GameCube game and built scripts that orchestrate the AI in the Assembly and Assembly matching, it is connected to the Ghidra CLI so it has access to that too. It’s impressive how good Opus is at that. If I have a empty .c file and the .o file of the original file, it decompiles the file to a very big percentage, based on size of the file. But decompilation is a long and very iterative process. It consumes tokens like a vacuum BUT since Copilot is based on Request rather than usage, I built scripts that will make the AI iterate over and over until it hits 100% Assembly matching. I once had over 30 Million Tokens consumed but it still costs me 3 premium requests, that’s a equivalent to one Opus Prompt. I don’t know if that will get „patched“ but decompilation hit a revolutionary level or efficiency
- Text
- verbatim
- Model
- Claude, Copilot
- Source
- Reddit ↗
- Published
- 2026-03
- Added
- 2026-09-26