이 에피소드에 관해
Almost every public Bitcoin node still runs Bitcoin Core. Brandon Black argues that client monoculture is a systemic risk — and that automation plus AI now make independent consensus implementations more realistic than skeptics assume.
Brandon (aka reardencode) returns to Stephan Livera Podcast for the first episode in a mini-series on non-Core Bitcoin implementations. He is shipping rbitcoin, a Rust full node aimed at server-side wallet and Lightning backends, with Electrum served in-process.
They dig into why rbitcoin has no UTXO set and no Core-style dbcache, how build-time differential testing differs from satd’s in-process libbitcoinconsensus dual-eval, BIP324 v2-only P2P, archival storage tradeoffs, LibreRelay-inspired policy, and what “production-ready™” means when every first-party line was written by AI under his prompting.
Brandon is blunt about maturity: the project is still early, contributors are welcome at rbitcoin.org, and listeners should treat alt clients as high-scrutiny infrastructure — not a drop-in replacement for Core tomorrow.
Timestamps:
00:00 — Intro: Brandon Black & rbitcoin
00:34 — Why Build rbitcoin
03:46 — Alt Clients Mini-Series
04:39 — Consensus Divergence Risk
05:53 — satd vs rbitcoin Paths
06:51 — Automation Makes Diversity Possible
08:29 — Build-Time Differential Testing
09:20 — Why Consensus Bugs Matter
10:55 — Lessons from btcd Divergences
13:35 — Target User: Wallet Backends
15:29 — Miners Leave the Non-Goal List
16:47 — Kill the dbcache Model
20:50 — Electrum In-Process
22:23 — Silent Payments & Tor
23:01 — BIP324 v2-Only P2P
23:46 — AI-Coded: Zero Human Lines
25:04 — Grok Plus Multi-Model Review
27:01 — DoS Rules & LibreRelay
28:34 — Upgrade Hooks
29:34 — Archive-Only, No 'UTXO Set'
31:50 — AI as the New Compiler
33:52 — Context Limits & Dependencies
38:06 — rust-bitcoin Relationship
39:08 — Wallet Infra Opportunity
40:00 — Still 10% a Joke
41:35 — Future of Client Diversity
44:50 — Outro: rbitcoin.org
Links:
Stephan Livera links:
Follow me on X: @stephanlivera
