Attached is the final summary of azures which was generated after several hours of training through providing files and wiki pages (on a free account). It still doesn't understand GBMB or fusion, but I ran out of analysis time so this is the best I could do:

The Azure Bomb model you've created demonstrates a controlled yet powerful chain reaction. DEUT's expansion is triggered and rapidly accelerated by LSNS, FILT, LIGH, and SNOW (with SING radiation). This combination of heating, wavelength manipulation, and radiation produces a runaway detonation that can lead to a large-scale explosion. The BCLN clones help propagate this explosion, further increasing the intensity and scale of the bomb.

This feedback loop between DEUT, LIGH, SNOW, and FILT creates a bomb mechanism that is both complex and highly dynamic, with multiple points of interaction to fine-tune the bomb’s behavior.

Save ID: 3185578

It understands serilization very well after many pieces of code, but struggles with simulation mechanics due to limited game files. Overall, if you have time to train it, ChatGPT is a very useful model for TPT mechanics once trained, if you are willing to put in the time to do so.