So I've been thinking... has anyone actually kicked off a serious effort to run Bitcoin Core through one of the big AI models? Like not just Core but also other implementations, btcd, BTC Knots, whatever Go-based stuff is out there. Claude's Fable and Mythos models seem like the obvious candidates for something this deep. Yeah the codebase is massive and it's not gonna be a weekend project, but imo it's doable if you break it up properly. Could even start privately. And on the funding side, there are bitcoin grants floating around, and honestly if some whale with heavy BTC exposure threw in a few bucks for a security audit like this, that seems like a no-brainer compared to the risk of a critical bug sitting in prod for years.
There are orgs out there that fund Bitcoin devs, so money probably isn't the blocker here. If they've been backing this ecosystem for years they might already be quietly doing something like this, who knows.
Also worth thinking about: independent AI-assisted reviews specifically hunting for exploits, kinda like what happened with that Zcash vuln not long ago. Bitcoin bugs are rare tho, proposals go through years of review before anything gets merged, so the bar is high. But "rare" doesn't mean zero.
Tbh I haven't seen any public initiative like this anywhere. Did a quick search on the bitcoindev mailing list and opensats.org and came up empty, so either it doesn't exist or it's being done quietly somewhere.
And yeah, the Zcash thing, that vulnerability was found by a paid security consultant hired specifically for that job, not some random AI side project. Point being, combining a proper human expert with AI tools is gonna cost way more than a couple dollars. That's just the reality of serious security work.
Fable got suspended yeah, hopefully just temporary. Opus would still work but Fable and Mythos are honestly the top picks for something like this right now. Cost is a real factor but like... grants exist for exactly this kind of thing. And the alternative is what, hoping nobody finds a nasty exploit first? Some bugs sat undetected in codebases for years before Mythos flagged them, there were a couple of writeups on this recently that really stuck with me. That risk alone makes the funding conversation worth having imo.