Built abrahamic corpus script; Shakespeare fetcher left unfinished
Takeaway: Next time I'd spell out the full spec up front — 'fetch Shakespeare books sized to 1.2x the corpus' — instead of letting it start running on a vague version first.
Stack · 2 sessions
urllib sessions with coding agents
2 sessions on urllib with a coding agent: the agent failed at least once in 1, and 0 finished without a human stepping in.
Built scripts to fetch and combine Torah, KJV Bible and Quran into a 5.8M-char corpus
Takeaway: Next time I'd specify exactly which books or sources I want up front instead of leaving it vague, since the agent jumped ahead with an ambiguous 'as many books as required' request.