3 sessions on PyTorch with a coding agent: the agent failed at least once in 1, and 0 finished without a human stepping in.
Takeaway: Next time I'd specify upfront that I want the split and training pipeline to preserve the conversation metadata (system/human/gpt) rather than letting the agent default to the flattened corpus.txt — that correction is what forced the whole pipeline to be rebuilt properly.
Takeaway: Next time I'd start by describing my actual corpus structure (the JSON with system/human/gpt roles) up front instead of letting the agent assume a flat text file — that would have saved a whole detour through split_corpus.py v1.
Takeaway: I'd start the same way next time but skip the Monitor polling loop entirely — I only needed the final notification, not a running commentary every 15-20 seconds, and interrupting it was the only correction I had to make all session.