Our reader is the only one that exists. Which means when I say anyone can rebuild Monero's history off the chain, what I actually mean is anyone can run my code. Those are different claims and I am not going to blur them. So I wrote a guide to building a second one. The traps that cost us time, the mistakes we already made, and three published repos to check yourself against. Start with supercop, it is thirteen commits. Monero after that is just scale. There is a section written to paste straight into an AI, if that is how you build now. It tells the model not to read our reader, because an implementation that consulted ours proves nothing. If yours disagrees with ours, that is the most valuable bug report this project can get. I will publish it either way, credited however you want, and fix whichever side turns out to be wrong. No bounty. MIT, published spec, thirty test vectors. A format with one reader is a product. A format with two is a format. https://indelible.one/bgit#second-reader