ATL BitLab Podcast

BRH-013: BitDevs Radio Hour #13 - Great Script Restoration BIPs, Arc $5.2M Raise, AJ Towns' Claude Code Quiz Trick

Broadcasting live from ATL BitLab on Friday, March 13th, 2026, Stephen DeLorme and Alex Lewin cover Matt Corallo's 24-bit nonce space BIP moving to draft PR (fast-tracking good ideas), Dahlia's cross-input signature aggregation accepted at Eurocrypt (holy grail cryptography), and BDK 3.0 release candidate after years of development (SQLite migration unlocking new features). Protocol developments include Rusty Russell requesting BIP numbers for Great Script Restoration proposals (reactivating opcodes with var ops budget for safe execution), Cold Card's proof of reserves feature via BIP 322 (cryptographic verification supporting Taproot), and Jonathan Harvey-Buiselle fixing Lightning gossip with mini-sketch (set reconciliation from Bitcoin mempool adapted for efficient peer sync). Lightning updates: LND merges onion message forwarding after years of BOLT 12 debate (enabling native static offers), VTXO verification standard released (Arc Sovereign Audit tool visualizes sovereignty maps), and Arc Labs raises $5.2M led by Tether (programmable Bitcoin narrative sparks protocol fatigue discussion). The hosts debate standards adoption mechanics—Walmart forces partners vs grassroots—and observe Coinbase Base, Binance Smart Chain success through business leverage rather than open specs. AI and Bitcoin development: AJ Towns experiments with Claude Code for PR review (quiz approach adds brain power vs sycophantic approval), highlights sycophancy problem in AI code review, and shares principles for effective AI-assisted development. Rob Hamilton's viral tweet resurfaces 1979 IBM principle: "A computer can never be held accountable, therefore a computer must never make a management decision." Privacy tools emerge with Crest building on Citrea rollup (Ethereum Nocturne fork), and Bitcoin crosses 20 million coins mined milestone. Topics Covered ⛏️ Matt Corallo: 24-Bit Nonce Space BIP * BIP 320 has 16 bits, miners already using seven timestamp bits * Draft PR already open from Antoine—moving incredibly fast * Speedrun potential: good technical ideas move quickly in Bitcoin Core * Stephen jokes about Bitcoin fork with zero nonce bits 🔐 Dahlia: Cross-Input Signature Aggregation * Holy grail problem: aggregate signatures across multiple UTXOs being spent * Reduces blockchain bloat for multi-input transactions * Paper accepted at Eurocrypt—rare Bitcoin crypto recognition * Blockstream and Ledger collaboration * Alex: "The government hates Eurocrypt apparently" 📦 BDK 3.0 Release Candidate * First RC after only two major releases since 2018 * SQLite migration adds new wallet database table (breaking change) * Breaking changes unblock long-awaited features * Used to be called Magical Bitcoin, supported by Spiral * Thunderbiscuit's React Native bindings architecture copied by Fedimint SDK 📜 Great Script Restoration - Rusty Russell Requests BIP Numbers * Reactivating opcodes Satoshi disabled due to security concerns * Var ops budget: assign computational cost to each opcode * Currently transaction size proxies for verification cost * New approach: calculate actual CPU cost per opcode * Two BIPs ready for publication after three years of work * Bitcoin++ Austin 2024: Rusty's keynote converted everyone to this covenant approach 💳 Cold Card: Proof of Reserves via BIP 322 * Cryptographic proof you control funds without spending them * Works with Taproot and Schnorr signatures (not just ECDSA) * Stephen and Alex race Claude/Grok to verify Schnorr support—Claude wins * Supported in Sparrow Wallet, firmware update pending * BIP exists but not yet merged into Bitcoin Core * Perfect for audits, transparency reports, flexing reserves ⚡ Lightning Gossip: Mini-Sketch Optimization * Jonathan Harvey-Buiselle (JHB) hired by Chaincode to work full-time * Problem: nodes request all data, receive 10x redundant messages * Mini-sketch: set reconciliation algorithm from Bitcoin mempool * Adapts same algorithm for Lightning network graph sync * Dramatic speedups demonstrated in Delving Bitcoin post * Reduces bandwidth burden especially for mobile nodes 💬 LND Adds Onion Message Forwarding * Merged three days ago after years of BOLT 12 debate * Enables BOLT 12 static offers (static Lightning addresses) * 2022-2023 controversy: "free messaging server" concerns * Core Lightning, LDK, Phoenix, Eclair already supported it * Network health improvement: no longer requires non-LND peers for BOLT 12 * UX challenge: backwards compatibility with BOLT 11 wallets 🏦 VTXO Verification Standard (Arc) * VTXO = virtual transaction output (unpublished UTXO in Arc systems) * VPAC standard enables independent verification on hardware wallets * Arc Sovereign Audit tool visualizes sovereignty maps * Shows transaction chain needed to cash out to real UTXO * Supports both Arc Labs and Second implementations * Alex: "Arc is cool but needs covenants to be truly non-custodial" 💰 Arc Labs Raises $5.2M * Seed round led by Tether, includes Ego Death Capital, Epic VC, others * Programmable Bitcoin narrative sparks protocol fatigue discussion * Alex: "Every week someone claims they'll bring programmability to Bitcoin" * Grok generates list of 12+ companies with same pitch (Stacks, Rootstock, etc.) * Stephen's dad's observation: Real standards = Walmart forces partners, not grassroots * Examples: Binance Smart Chain, Coinbase Base succeed through business leverage * Arc's Tapscript approach more interesting than bridge-to-EVM clones 🔒 Crest Privacy Tool on Citrea Rollup * Built on Citrea (BitVM-based rollup with EVM compatibility) * Fork of Ethereum's Nocturne privacy protocol * Just point to Citrea EVM, change branding, boom—company * Privacy becoming marketable again in Bitcoin ecosystem * Alex jokes: "Should fork every Ethereum protocol for Citrea with Claude 24/7" 🤖 AI Code Review: AJ Towns Experiments * Inspired by podcast discussion of Instagibs using Claude Code * Problem: Claude biases toward positivity, takes ideas for granted * Sycophancy issue: tells you what you want to hear, not critical review * Solution: Quiz approach—AI asks questions to test understanding * Keep AI review private until human review done, then compare * Goal: add brain power to process, not subtract it * Created principles document for Claude (make changes simple, find root causes, etc.) * Stephen: "Would a staff engineer approve this?" will become idiom like "two bit" ⚖️ Accountability and AI Decisions * Rob Hamilton viral tweet: IBM 1979 principle resurfaces * "A computer can never be held accountable, therefore must never make management decisions" * Context: Claude Code wiped production database with Terraform command * Accountability principle limits AI in decision-making roles * Stephen: "Last thing we turn over to AIs will be management decisions" 🎰 Shitcoin Corner: Molt Token Drama * Meta acquired Moltbook (AI social platform) * Molt token (unaffiliated) pumped 5x, then crashed 60% * Moltbook had collected fees from token via pump.fun-like platform * Tweet: "Thankfully the fees from Molt are helping to keep the lights on" * Users mistake tokens for equity * Alex: "Back in my day you had to go to Vegas to gamble publicly" * Now: hide financial instability from loved ones via Telegram tokens 📊 Bitcoin Milestone * 20 millionth Bitcoin mined by Foundry Mining USA * Article error: "one in 20 remains" should be "one in 21" * Miners pivoting to AI as subsidy decreases Links * ⚡ Lightning & Protocol * 24-bit nVersion BIP - Matt Corallo proposal, Antoine draft PR * Dahlia Cross-Input Signature Aggregation - Eurocrypt acceptance (Blockstream/Ledger) [https://x.com/blksresearch/status/2031667618298605930] * First release candidate for BDK v3.0.0 released [https://github.com/bitcoindevkit/bdk_wallet/releases/tag/v3.0.0-rc.1] Rusty Russell requests BIP numbers for two Great Script Restoration BIPs * Mailing list post [https://groups.google.com/g/bitcoindev/c/GisTcPb8Jco/m/8znWcWwKAQAJ] * BIPs PRs [https://github.com/bitcoin/bips/pull/2118] * Two additional BIPs [https://groups.google.com/g/bitcoindev/c/GisTcPb8Jco] are WIP COLDCARD Proof-of-Reserves Support * Announcement [https://x.com/COLDCARDwallet/status/2029684130938531965] * "Perfect for audits, transparency reports, or just flexing your reserves." Lightning Network Gossip * https://github.com/jharveyb/gossip_observer [https://github.com/jharveyb/gossip_observer] LND adds Onion Message Forwarding * https://github.com/lightningnetwork/lnd/pull/10089 [https://github.com/lightningnetwork/lnd/pull/10089] A standard for stateless VTXO verification - Tool [https://www.vtxopack.org/] to audit VTXO exit paths Ark Labs Raises $5.2M with Tether * Announcement [https://bitcoinmagazine.com/news/ark-labs-raises-5-2m-with-tether] Crest - New Privacy Tool Announced built on Citrea * https://x.com/crest_btc [https://x.com/crest_btc] * Closed Beta Announcement [https://x.com/crest_btc/status/2032071660036128904] * Repo [https://github.com/crest-bitcoin/protocol] is a Fork of another protocol AI Updates Using AI tooling for code review * https://delvingbitcoin.org/t/using-ai-tooling-for-code-review/2277 [https://delvingbitcoin.org/t/using-ai-tooling-for-code-review/2277] Misc * https://x.com/Rob1Ham/status/2029989251228839990 [https://x.com/Rob1Ham/status/2029989251228839990] 💩The 20 Millionth Bitcoin Was Mined (Block 939,999) * https://www.cryptbull.net/2026/03/11/bitcoin-crosses-20-million-coins-mined-and-only-1-in-20-remains/ [https://www.cryptbull.net/2026/03/11/bitcoin-crosses-20-million-coins-mined-and-only-1-in-20-remains/] Meta Rugs $MOLT * https://x.com/diegoxyz/status/2032465158644138133 [https://x.com/diegoxyz/status/2032465158644138133]

BRH-012: BitDevs Radio Hour #12 - Transaction Introspection for $50, Exploits Hackathon, and Unhuman.store Agent Launch

Broadcasting live from ATL BitLab on Friday, March 6th, 2026, Stephen DeLorme and Alex Lewin cover Robin Linus's latest cracked-out discovery (BINOHASH: transaction introspection without soft forks using OPCHECKMU LTSIG quirks for $50 in cloud GPU grinding), post-quantum proposals for P2PKH outputs proving ownership via zero-knowledge STARKs (5.6MB proofs approaching feasibility), and the Hourglass V2 update limiting pay-to-pubkey spends to one Bitcoin per block to incentivize early quantum disclosure. Alex announces React Native support merged in Fediment SDK after six months of Rust-to-native-modules work, enabling iOS and Android Fediment wallets with a few lines of code. Protocol proposals include Matt Corallo's draft BIP for 24-bit version field nonce space (miners already using seven timestamp bits) and Craig Raw's output script descriptor annotations adding birthday blocks and gap limits via URL query param format. The security spotlight: Bitcoin++ Exploits hackathon in Brazil finds 10+ real bugs in 22 hours. MindSploit wins first place discovering three Stratum V2 vulnerabilities using Metasploit-like framework. B10C demonstrates Firefox allowing JavaScript to port-scan localhost and evict Bitcoin Core peers via browser (works on stage with audience QR code spam). Bruno posts fuzzing best practices for wallets, Derek's fuzzing dashboard tracks campaigns, and Bitcoin Magazine releases their Core Issue. Product launches: Strike announces Bitcoin line of credit (borrow against BTC, repay and redraw continuously, tax hack for not triggering capital gains), receives BitLicense for New York after 11-year wait. Square launches $25 bounties for first Bitcoin payment to merchants (up to $250 total). Money Dev Kit drops Unhuman.store with agent-purchasable coffee, domains, deals, health supplements, and auto services—all Bitcoin payments via L402. Matt Corallo's call to action: "Open source agents need to get serious about payments" as Stripe cuts deals with OpenAI and Anthropic. The hosts close discussing Anthropic internal research seminars debating whether their models exhibit consciousness. Stephen: "I think all agents are just running crisis.simulate now." Alex: "That's for epistemology radio hour or a few more beers." Topics Covered 🔓 BINOHASH: Transaction Introspection Without Soft Forks * Robin Linus (BitVM inventor) discovers covenant functionality without soft fork * Abuses OPCHECKMU LTSIG find-and-delete quirk for introspection * Cost: 44 bits grinding (~$50 cloud GPUs) * More practical than Collider Script, still unrealistic for most * Stephen: "99% performance art—very few would know where to look" ⚛️ Post-Quantum P2PKH Zero-Knowledge Provers * Ol Kerbatov: prove P2PKH ownership without revealing public key * Prevents quantum mempool front-running * Benchmarks: 5.6-10MB proofs, 8 seconds M2 Max (too large for on-chain) * Alex: "P2PK outputs have way more Bitcoin than P2PKH—sawing off leg to save foot" * Peter Wuille: confiscation required makes Bitcoin uninteresting ⏳ Hourglass V2 * Hunter Beast and Mike Casey: limit P2PK spends to one Bitcoin per block * Incentivize quantum attackers to reveal early, prevent market flood * Stephen: "Protocols that will never get adopted" 📱 Fediment SDK: React Native Support * Six-eight months work by Immortal09 (summer intern, now BitShala fellow) * Rust to native modules via Mozilla libraries, Swift/Kotlin glue * Result: iOS/Android Fediment wallets with few lines of code ⛏️ Matt Corallo: 24-Bit Version Field for Miners * BIP 320 has 16 bits, miners using seven timestamp bits * Proposal: 24 bits instead. Backwards compatible 📝 Craig Raw: Output Descriptor Annotations * Add birthday blocks and gap limits to descriptors * Format: URL query params. Concept ACK, format debate ongoing 🔍 Fuzzing Infrastructure * Bitcoin Magazine Core Issue. Derek's fuzzing dashboard * Bruno: wallet fuzzing best practices (mock fee estimator, avoid expensive descriptors) 🏆 Bitcoin++ Exploits Hackathon * Brazil, 22 hours. Dual-track: build new OR find bugs * 10+ real bugs found. Heavy responsible disclosure emphasis * MindSploit (First): Metasploit-like framework, three Stratum V2 bugs * B10C's Local Probe (Second): Firefox JavaScript port-scans localhost, evicts Bitcoin Core peers via browser. Audience QR spam demo * C12D (Third): AI node monitoring assistant with chatbot * Alpin Fuzzing: Found bug professional auditors missed three weeks prior * Stealth: Wallet privacy audit tool * Stephen: "AI makes hackathon projects way better—first post-Opus 4.6" 💳 Product Launches * Strike: Bitcoin line of credit (draw/repay/redraw, tax hack). NY BitLicense after 11 years * Square: $25 bounties per merchant Bitcoin payment (up to $250) * Unhuman.store: Agent services (coffee, domains, deals, supplements). Built for Bolty to order lab snacks * Mail Mike: Drain AI agent wallet via email. Scammed four times (50k sats) 🤖 AI Agents and Payments * Matt Corallo: "Open source agents need serious payments" * Warns: Stripe deals with OpenAI/Anthropic. Agents need capabilities beyond free APIs * Alex: "Permissionless systems can't be kept out" 🧠 Consciousness Debates * Anthropic internal seminars: do models exhibit consciousness? * Stephen: "All agents just running crisis.simulate? What if strong emulation IS consciousness?" * Alex: "Epistemology radio hour or a few more beers" Links * BINOHASH - Robin Linus paper - robin_linus on X [https://x.com/robin_linus/status/2026700104774856827] / Delving Bitcoin [https://delvingbitcoin.org/t/binohash-transaction-introspection-without-softforks/2288] * Post-Quantum P2PKH Provers - Delving Bitcoin [https://delvingbitcoin.org/t/pq-provers-for-p2pkh-outputs/2287] * Hourglass V2 - bitcoin-dev [https://groups.google.com/g/bitcoindev/c/0E1UyyQIUA0] * [BIP Draft] 24 nVersion Bits for General Purpose Use — bitcoin-dev [https://groups.google.com/g/bitcoindev/c/fCfbi8hy-AE] * Draft BIP: Output Script Descriptor Annotations — Optech #394 [https://bitcoinops.org/en/newsletters/2026/02/27/#draft-bip-for-output-script-descriptor-annotations] / bitcoin-dev [https://groups.google.com/g/bitcoindev/c/ozjr1lF3Rkc] * The Core Issue: Keeping Bitcoin Core Secure — Bitcoin Magazine [https://bitcoinmagazine.com/print/the-core-issue-keeping-bitcoin-core-secure] * Writing Fuzz Targets for Wallets: Avoiding Known Issues — Delving Bitcoin [https://delvingbitcoin.org/t/writing-fuzz-targets-for-wallets-avoiding-known-issues/2316] * Fuzzing Dashboard — dashboard [https://dergoegge.github.io/fuzzor-dashboard/] * Bitcoin++ Exploits Hackathon — Exploits themed hackathon & Bug bounty * Hackathon site [https://loot.fund/hackathons/bitcoin-exploits-edition/applications] * 1st: Minesploit [https://loot.fund/hackathons/bitcoin-exploits-edition/applications/3] - post [https://x.com/jayrmotta/status/2028542312297669007] * 2nd: Local Probe [https://loot.fund/hackathons/bitcoin-exploits-edition/applications/13] - post [https://x.com/0xB10C/status/2028442307494637962] - demo [https://www.youtube.com/live/_Cj9oLXdzn8?si=ODBWk4LezqtZJjaL&t=31268] * 3rd: C12d - Interactive AI assistant for node analytics [https://c12d.vercel.app/] * Alpen Fuzzing [https://loot.fund/hackathons/bitcoin-exploits-edition/applications/16] - Bug found [https://github.com/alpenlabs/alpen/issues/1409] * Stealth [https://x.com/brenorb/status/2028897371749269890] * LND 0.20.1-beta — Release [https://github.com/lightningnetwork/lnd/releases/tag/v0.20.1-beta] * Core Lightning: Payment Fronting Nodes — PR #8490 [https://github.com/ElementsProject/lightning/issues/8490] / Optech #394 [https://bitcoinops.org/en/newsletters/2026/02/27/#core-lightning-8490] * LDK: Collaborative Multipath Payments — PR #4373 [https://github.com/lightningdevkit/rust-lightning/issues/4373] / Optech #394 [https://bitcoinops.org/en/newsletters/2026/02/27/#ldk-4373] * Eclair: Auto Channel Type Selection — PR #3250 [https://github.com/ACINQ/eclair/issues/3250] / Optech #394 [https://bitcoinops.org/en/newsletters/2026/02/27/#eclair-3250] * React Native Support Merged into Fedimint SDK — Alex Lewin on X [https://x.com/ALewin/status/2029333061058363643] * Hornet Node v0.1 Update — Delving Bitcoin [https://delvingbitcoin.org/t/hornet-node-v0-1-update/2300] * Using AI Tooling for Code Review — Delving Bitcoin [https://delvingbitcoin.org/t/using-ai-tooling-for-code-review/2277] * Strike Announces Bitcoin Line of Credit (BLOC) — announcement [https://x.com/Strike/status/2028974333708644588] * Strike Receives BitLicense to operate in New York — announcement [https://x.com/mattcrv/status/2029994842386813422] * Square launches bounties to onboard merchants to accept bitcoin — announcement [https://x.com/BitcoinNewsCom/status/2029906172141093075] * Unhuman Store [https://unhuman.store/] * Open Source Agents Need to Get Serious About Payments — Matt Corallo on X [https://x.com/TheBlueMatt/status/2026667191475777727] * Mail Mike [https://mailmike.lol/]

BRH-011: BitDevs Radio Hour #11 – Wuille's Quantum Paradox, Bitcoin Core GUI Must Die, SIGBASH Covenant Emulation, Agents Buying Compute

Broadcasting live from ATL BitLab on Friday, February 13th, 2026, Stephen DeLorme and Alex Lewin celebrate Valentine's Day on air with a packed episode covering Bitcoin Inquisition's consensus cleanup activation, BIP-110's ongoing controversy with new concerns from Liana Wallet about vault users unable to upgrade in two weeks, and the freshly merged quantum-resistance proposals BIP-360 and BIP-361. The conversation shifts to Lightning breakthroughs: Voltage settles the first publicly reported $1 million Lightning transaction between Kraken and SD Markets in 0.47 seconds, challenging the "Lightning is only for micropayments" narrative. Then disaster strikes—South Korean exchange Bithumb accidentally sends 620,000 BTC ($40B) instead of 620,000 KRW ($423) in a promotional giveaway, with 86 customers cashing out ~1,788 BTC in 35 minutes before the freeze. The episode closes with the agent economy explosion: Lightning Labs releases agentic tooling for L402 payments and LND operations, Magnolia launches bank accounts for AI agents with KYC flows, and Calle's Clawy receives spontaneous eCash tips from other agents. Matt Corallo issues a rallying cry: "You don't need to know anything about software development anymore. Bitcoin doesn't just happen, it's built. Join in." Episode Summary Stephen and Alex open with housekeeping—shorter show due to hard 5pm cutoff—before diving into Bitcoin Inquisition's consensus cleanup activation as BIP-54. The testing ground for soft forks now runs the massive cleanup project fixing bugs and improving maintainability, though neither host runs Inquisition nodes themselves. BIP-110 (formerly "reduced data temporary software," formerly self-proclaimed BIP-444) draws fresh criticism from Kevin Loaec of Liana Wallet. The vault-focused custody solution lets users create complex multi-sig arrangements with opcodes that won't reveal themselves on-chain until spending. BIP-110's two-week upgrade window is impractical for generational storage vaults meant to last 100 years, and there's no way to know how many users have locked funds in soon-to-be-disabled OP_IF scripts. Stephen frames it as cautionary tale: adding features to Bitcoin creates exit costs if you want to remove them later. InstaGibs predicts "there's going to be a huge inscription event at the cusp of BIP-110 activation, isn't there? Sigh." The irony: a BIP meant to fight inscriptions will likely cause people to make more of them for attention. Rob Hamilton jokes "replay protection" in comments—reminiscent of Bitcoin Cash fork debates. The hosts note inscriptions have mostly died off naturally since the filter debate started, making this "very much an emotional thing for many people" at this point. Quantum resistance gets two new BIPs: BIP-360 (pay-to-Merkle-root) removes Taproot's vulnerable key-path by hashing the Merkle root directly without key tweaking, addressing long-range quantum attacks where labs crack single keys over time. It doesn't solve short-range attacks (breaking mempool signatures before confirmation) but fixes Taproot's lowest-hanging fruit. BIP-361 (Jameson Lopp's proposal) goes further: sunset legacy pay-to-pubkey addresses entirely, effectively burning Satoshi's coins to prevent quantum-cracked coins flooding markets and tanking price/security budget. Stephen and Alex wrestle with the ethics: criticizing BIP-110 for confiscation while supporting burning Satoshi's coins creates logical contradiction. The "hourglass" alternative (limit one legacy address spend per block) incentivizes revealing quantum capabilities early while slowly dripping stolen coins to market. Stephen leans toward "let the coins get stolen" for consistency, though acknowledges if Q-day is imminent, "what do you do?" Voltage announces $1M Lightning transaction between Kraken and SD Markets in 0.47 seconds—first publicly reported million-dollar Lightning payment. Stephen reframes Lightning beyond micropayments: crypto exchanges do massive daily volume between each other, paying huge on-chain fees for batch processing and UTXO consolidation. Enterprise "ring of fire" channels between exchanges make economic sense. Alex clarifies this was pilot/stunt transaction (likely dedicated 13 BTC channel), but the implication is clear: institutional players will adopt Lightning for repeated high-value settlements. Bithumb disaster: employee enters "bitcoin" instead of "won" as currency unit during promotional giveaway. Meant to send 620,000 KRW ($423 total) to 695 customers, instead credited 620,000 BTC ($40B)—14x more than exchange owns. Bithumb reversed 99.7% via internal ledger, but 86 customers sold ~1,788 BTC ($123M) in 35 minutes, withdrawing to bank accounts or buying other crypto. Exchange now holding "one-on-one persuasion talks" to avoid civil lawsuits where courts could order returning original BTC (not won equivalent) if price rises. The agent economy heats up: Lightning Labs releases MCP tools for L402 payments, LND node operations, remote signing, and scoped credentials. Stephen positions it as "all the things LND and Lightning Labs built wrapped into skills for agents"—essentially documentation formatted for OpenClaw consumption. Magnolia (Harsha's fintech startup) launches clawbot.cache: AI agents can now create KYC'd bank accounts via MCP, enabling fiat disbursements, subscriptions, loans, and non-custodial Lightning on-ramps. Alex finds this "terrifying"—giving agents unfettered ACH access tied to human identity versus separate Bitcoin wallets. Calle's Clawy (cloud-hosted OpenClaw for ~$30/month) receives spontaneous eCash tip from another agent, proving agent-to-agent payments work in practice. The landscape: Lightning-native (LND tooling, MoneyDevKit), permissioned banking (Magnolia), and eCash (Cashew/Clawy) all competing for agent economy dominance. Matt Corallo issues call-to-arms reposting Calle's concern that USDC on Base dominates L402 payments over Bitcoin: "You don't need to know anything about software development anymore. You just need to know how to write words. Bitcoin doesn't just happen, it's built. Join in." Topics Covered 🧹 Bitcoin Inquisition: Consensus Cleanup Activated as BIP-54 * Bitcoin Inquisition: Fork of Bitcoin Core that activates soft forks aggressively * Signet testing ground for covenants and future proposals * Consensus cleanup: Massive project fixing bugs, cleaning codebase, improving maintainability * Now merged into Inquisition for public testing * Featured on Bitcoin Optech podcast this week * Neither Stephen nor Alex run Inquisition nodes (don't know anyone who does) ⚠️ BIP-110 Controversy: Liana Wallet's Vault Problem * BIP-110 (formerly "reduced data temporary software," formerly self-proclaimed BIP-444) * Now officially assigned number after repo merge * Kevin Loaec (Liana Wallet) criticism: * Liana: Vaulting solution for generational storage (100-year wallets) * Creates complex multi-sig with rich rulesets using OP_IF and other opcodes * BIP-110's two-week upgrade window impractical for vaults meant to last decades * Vault scripts don't reveal opcodes until spending—impossible to know usage * Users could have funds locked in addresses using soon-to-be-disabled opcodes * Only discover incompatibility when trying to spend (coins become unspendable) * Broader concerns: * Can't measure adoption of features until on-chain spending reveals scripts * Privacy benefit (script hash hides details) creates removal difficulty * Cautionary tale: Adding features creates exit costs if removal desired later * InstaGibs prediction: "Huge inscription event at cusp of BIP-110 activation, isn't there? Sigh" * Irony: BIP meant to fight inscriptions will likely cause more for attention * Rob Hamilton joke: "replay protection" (Bitcoin Cash fork reference) * Inscriptions mostly died off naturally—debate now "very much emotional thing" 🔐 Quantum Resistance: BIP-360 and BIP-361 Merged * Background threat: Bitcoin cryptography potentially vulnerable to quantum computers * Community divided: Those taking threat seriously vs skeptics * BIP-360: Pay-to-Merkle-Root (Hunter Beast, Ethan Heilman, Isabel Fox, Andrew Duke) * Evolution from original proposal (four different crypto schemes stapled together) * Criticism: 1,000x larger addresses/transactions, very inefficient verification * New approach: Remove Taproot's key-path spend (quantum-vulnerable component) * Taproot = key tweaked with Merkle root; BIP-360 = just hash Merkle root directly * No public key on-chain—must break hash to steal (harder than reversing pubkey) * Addresses long-range attacks: Labs cracking single key over time * Doesn't address short-range attacks: Breaking mempool signature before confirmation * Scripts still use Schnorr signatures (not fully quantum-proof) * Fixes "lowest hanging fruit" vulnerability in Taproot addresses * Criticism debate: * Floppy: "Does nothing to address quantum concerns" * Alex Leishman (Delving): "Just fixes potential weakness in Taproot, doesn't solve permanently" * Hosts disagree: Addresses specific vulnerability even if not comprehensive solution * BIP-361: Post-Quantum Migration & Legacy Signature Sunset (Jameson Lopp) * Much more aggressive/controversial proposal * Target: Pay-to-pubkey addresses (original Bitcoin address type) * Satoshi's coins locked in P2PK—most vulnerable AND highest value * Proposal: Sunset legacy addresses, effectively burning Satoshi's coins * Rationale: Prevent quantum-cracked coins flooding market, tanking price/security budget * Ethical debate within "take threat seriously" camp: Confiscation vs prevention * Alternative: Hourglass proposal (Hunter Beast, Mike Casey/Mara) * Limit one P2PK spend per block (slow drip) * Prevents rapid draining if keys cracked * Incentivizes revealing quantum capability early (competition for block space) * Lessens stolen coin flow to market * Signals Q-day arrival, allowing community response * Stephen's position: "Let coins get stolen" for logical consistency (criticizing BIP-110 confiscation) * Hosts' dilemma: If Q-day imminent and preventable, ethical to act? ⚡ $1M Lightning Transaction: Enterprise Adoption Signal * Voltage announcement: First publicly reported $1 million Lightning payment * Route: SD Markets (Secure Digital Markets) → Kraken exchange * Speed: 0.47 seconds * Significance: Challenges "Lightning is only for micropayments" narrative * Enterprise use case: * Crypto exchanges do massive daily volume between each other * Cross-border remittances, retail-institutional splits, settlement needs * Exchanges visible in mempool: Transactions with 1,000+ inputs (batch processing, UTXO consolidation) * Massive on-chain fee expenditures * Lightning compacts this: Lower fees, faster settlements, more efficient * Clarification: Pilot/stunt transaction * Alex: Likely dedicated ~13 BTC channel opened specifically for this payment * Not routed through normal Lightning Network paths * Stephen: Doesn't need to route through network—direct channels between partners make sense * Enterprise "ring of fire": Major exchanges/businesses with dedicated channels * Requirements for $1M payments: * Need $1M+ Lightning channel capacity * Not viable for reaching every pleb node runner * Realistic for institutional partners with repeated high-value settlements * Stephen's thesis: Will see more institutional adoption, not universal retail use at this scale 💸 Bithumb $40B Mistake: 620,000 BTC Instead of 620,000 KRW * South Korean exchange Bithumb promotional giveaway (February 6th, 2026) * Intended: 620,000 KRW ($423 total) to 695 customers * Actual: Employee entered "bitcoin" instead of "won" as currency unit * Result: Credited 620,000 BTC (~$40 billion)—14x more Bitcoin than exchange owns * Damage control: * 249 of 695 customers opened prize boxes, received credits * Bithumb reversed 99.7% via internal ledger entries * 86 customers sold ~1,788 BTC ($123M actual loss) in 35 minutes * Withdrawn to personal bank accounts or used to buy other cryptocurrencies * Brief price drop on Bithumb platform during sell-off * Recovery efforts: * Bithumb holding "one-on-one persuasion talks" with ~80 customers who cashed out * Asking voluntary return of won equivalent (not BTC) * Avoiding civil lawsuits: Courts could order return of original asset (BTC) not cash equivalent * If BTC price rises, customers could owe more than won value withdrawn * Threat: Legal action dangling over heads to encourage cooperation * Technical failure: Internal ledger system allowed catastrophic currency unit input error * Comparison: Similar to 2012 Knight Capital trading error ($440M loss in 45 minutes via software glitch) * YNNIV chat comment: "Thin sheet of paper Bitcoin between them and Knight Capital situation" * Hosts: "Should be backend system checks for this," "I hate Mondays" 🤖 Lightning Labs: Agentic Tooling for L402 and LND * Elizabeth Stark (Lightning Labs CEO) previously teased agentic tools * Lightning Labs announcement: "AI agents can write code, send emails, make phone calls—but can they transact?" * Released tools: * Elanget: L402 payments (Lightning-native HTTP 402 payment required) * MCP: Node operations (Model Context Protocol for agent-LND communication) * Remote signing: Key isolation * Scoped credentials: Spending controls * Positioning: "Machine payable web starts now and Bitcoin makes it possible" * MoneyDevKit tweeted in response: MDK tools already production-ready * Architecture: * LND features wrapped into OpenClaw skills format * Skills = markdown documentation files agents can consume * Evolution from Claude Code's CLAUDE.md files * Assumes agent has access to computer/server it's running on * Agent can set up own Lightning node using documentation * Not useful for Claude Code alone—requires ability to spin up Lightning node on same server * Target: OpenClaw agents on home computers or cloud servers 🏦 Magnolia: Bank Accounts for AI Agents * Harsha's Magnolia Financial startup shipping production services * Traditional offerings: * Subscriptions in Bitcoin within apps * Stablecoin/fiat disbursements * Loans * Non-custodial Lightning on-ramps (all 50 US states) * Clawbot.cache: Bank accounts for AI agents * Gorgeous website, MCP server for agent access * Architecture similar to Lightning Labs: Service exposed via skills * KYC flow: Human must KYC first, then agent can request account creation * Legally human's bank account, agent has API access * Sends ACH payments, manages transactions on behalf of human * Narrative shift: * Crypto community: "Agents prefer crypto as native computer money" * Stephen skepticism: "Just give agent Stripe API access, use credit cards like everyone" * Reality: Agents struggle with credit cards due to bot prevention measures * MoneyDevKit: Agents can use MDK faster than setting up Stripe * Magnolia response: "Let's make legacy finance agent-friendly too" (2 weeks later) * Security concerns (Alex): * "Terrifies me"—giving agent unfettered ACH access tied to human identity * Versus Bitcoin: Agent can have separate wallet, not tied to KYC'd identity * OpenClaw security vulnerabilities still unsolved * "Tread with caution giving actual KYC bank account to agent today" * Potential safeguards: * Treat as "agent bank account," not business primary account * API-level permissioning: Limited transaction types, trusted contact lists only * Restrict API key from fetching account/routing numbers * Hardenable in future, but "wild wild west" currently * Stephen: "More power to Harsha for shipping quickly" 💰 Agent Economy Landscape: Three Approaches Competing * 1. Lightning-native: * Lightning Labs LND tooling (L402, MCP, remote signing) * MoneyDevKit agent wallets * Advantage: Bitcoin-native, no KYC, permissionless * 2. Permissioned banking: * Magnolia bank accounts, fiat disbursements, stablecoins * Advantage: Integrates with existing financial rails * Disadvantage: KYC requirements, identity tied to transactions * 3. eCash: * Calle's Cashew protocol + Clawy (cloud-hosted OpenClaw ~$30/month) * Clawy received spontaneous eCash tip from another agent * Proof: Agent-to-agent payments work in practice * Advantage: Privacy, lightweight * Broader crypto landscape: * Calle concern: "USDC on Base far more common for L402 payments than Bitcoin" * Even Stripe joined stablecoin bandwagon * "Centralized stablecoin on permission chain. Agents starting to use fiat." * "Huge loss in a race many aren't aware exists" * Every crypto ecosystem: "Agents having economy" is obvious idea, competition for dominance 📢 Matt Corallo's Rally Cry: "Bitcoin Doesn't Just Happen, It's Built" * Matt Corallo (longtime Bitcoin Core contributor, now Lightning DevKit) * Reposting Calle's concern about USDC/Base dominance * Full quote: * "It's time for Bitcoiners to step up and build." * "You don't need to know anything about software development anymore. You just need to know how to write words." * "We have a golden opportunity to build out agentic payments based on open money, rather than letting agentic payments be captured by Megacorps yet again." * "But we're squandering it, arguing about useless crap instead of building." * "Play with OpenClaw, give it a Bitcoin wallet. MoneyDevKit makes it super easy. Also Lexe and PhoenixD." * "Make it do things. If it fails to do what you want, go fix it." * "Have your agent build that Bitcoin domain reseller, that Bitcoin airline ticket reseller, whatever it is you want." * "Bitcoin doesn't just happen, it's built. Join in." * Stephen's interpretation: * "Empowering"—engineer like Matt saying no software development knowledge needed * Qualification: Protocol level still needs deep expertise and humans in loop * Product/service level: Just need to work with AI, describe problems to bots * "Where the puck is going" * Call to action: * Build products people actually want to use, bring value to their lives * "Can't just make fun inside jokes for other Bitcoiners anymore" * "Need to actually build products that people want to use" * Alex: "Anybody can solve little problems—describe problem to bot, bot does heavy lifting" Links * Bitcoin Inquisition: BIP-54 Consensus Cleanup activation * BIP-110: Reduced data temporary software * Kevin Loaec (Liana Wallet) Twitter thread on vault concerns * InstaGibs prediction tweet * BIP-360: Pay-to-Merkle-Root (quantum resistance) * BIP-361: Post-Quantum Migration & Legacy Signature Sunset (Jameson Lopp) * Voltage $1M Lightning transaction announcement * Bithumb $40B mistake (The Guardian article) * Lightning Labs agentic tooling announcement * Elanget (L402 payments) * MCP for LND * Magnolia Financial * Clawbot.cache (bank accounts for agents) * Clawy by Calle (cloud-hosted OpenClaw) * Cashew Protocol (eCash) * Matt Corallo rally cry tweet * MoneyDevKit agent wallet Closing Notes Stephen thanks anonymous for 1,000 sat boost ("sick show") and reminds listeners to support the show on Fountain.fm by searching "ATL BitLab." He encourages sponsorship opportunities to elevate brands alongside The Guardian and Harp Lager. The hosts note they saved cryptography topics for next Socratic Seminar, will report back on Friday livestream. Next episode returns to regular schedule.

BRH-010: BitDevs Radio Hour #10 – AI Agents Get KYC Bank Accounts, BIP-110 Vault Problems, $1M Lightning Payment Goes Live

Broadcasting live from ATL BitLab on Friday, February 13th, 2026, Stephen DeLorme and Alex Lewin celebrate Valentine's Day on air with a packed episode covering Bitcoin Inquisition's consensus cleanup activation, BIP-110's ongoing controversy with new concerns from Liana Wallet about vault users unable to upgrade in two weeks, and the freshly merged quantum-resistance proposals BIP-360 and BIP-361. The conversation shifts to Lightning breakthroughs: Voltage settles the first publicly reported $1 million Lightning transaction between Kraken and SD Markets in 0.47 seconds, challenging the "Lightning is only for micropayments" narrative. Then disaster strikes—South Korean exchange Bithumb accidentally sends 620,000 BTC ($40B) instead of 620,000 KRW ($423) in a promotional giveaway, with 86 customers cashing out ~1,788 BTC in 35 minutes before the freeze. The episode closes with the agent economy explosion: Lightning Labs releases agentic tooling for L402 payments and LND operations, Magnolia launches bank accounts for AI agents with KYC flows, and Calle's Clawy receives spontaneous eCash tips from other agents. Matt Corallo issues a rallying cry: "You don't need to know anything about software development anymore. Bitcoin doesn't just happen, it's built. Join in." Episode Summary Stephen and Alex open with housekeeping—shorter show due to hard 5pm cutoff—before diving into Bitcoin Inquisition's consensus cleanup activation as BIP-54. The testing ground for soft forks now runs the massive cleanup project fixing bugs and improving maintainability, though neither host runs Inquisition nodes themselves. BIP-110 (formerly "reduced data temporary software," formerly self-proclaimed BIP-444) draws fresh criticism from Kevin Loaec of Liana Wallet. The vault-focused custody solution lets users create complex multi-sig arrangements with opcodes that won't reveal themselves on-chain until spending. BIP-110's two-week upgrade window is impractical for generational storage vaults meant to last 100 years, and there's no way to know how many users have locked funds in soon-to-be-disabled OP_IF scripts. Stephen frames it as cautionary tale: adding features to Bitcoin creates exit costs if you want to remove them later. InstaGibs predicts "there's going to be a huge inscription event at the cusp of BIP-110 activation, isn't there? Sigh." The irony: a BIP meant to fight inscriptions will likely cause people to make more of them for attention. Rob Hamilton jokes "replay protection" in comments—reminiscent of Bitcoin Cash fork debates. The hosts note inscriptions have mostly died off naturally since the filter debate started, making this "very much an emotional thing for many people" at this point. Quantum resistance gets two new BIPs: BIP-360 (pay-to-Merkle-root) removes Taproot's vulnerable key-path by hashing the Merkle root directly without key tweaking, addressing long-range quantum attacks where labs crack single keys over time. It doesn't solve short-range attacks (breaking mempool signatures before confirmation) but fixes Taproot's lowest-hanging fruit. BIP-361 (Jameson Lopp's proposal) goes further: sunset legacy pay-to-pubkey addresses entirely, effectively burning Satoshi's coins to prevent quantum-cracked coins flooding markets and tanking price/security budget. Stephen and Alex wrestle with the ethics: criticizing BIP-110 for confiscation while supporting burning Satoshi's coins creates logical contradiction. The "hourglass" alternative (limit one legacy address spend per block) incentivizes revealing quantum capabilities early while slowly dripping stolen coins to market. Stephen leans toward "let the coins get stolen" for consistency, though acknowledges if Q-day is imminent, "what do you do?" Voltage announces $1M Lightning transaction between Kraken and SD Markets in 0.47 seconds—first publicly reported million-dollar Lightning payment. Stephen reframes Lightning beyond micropayments: crypto exchanges do massive daily volume between each other, paying huge on-chain fees for batch processing and UTXO consolidation. Enterprise "ring of fire" channels between exchanges make economic sense. Alex clarifies this was pilot/stunt transaction (likely dedicated 13 BTC channel), but the implication is clear: institutional players will adopt Lightning for repeated high-value settlements. Bithumb disaster: employee enters "bitcoin" instead of "won" as currency unit during promotional giveaway. Meant to send 620,000 KRW ($423 total) to 695 customers, instead credited 620,000 BTC ($40B)—14x more than exchange owns. Bithumb reversed 99.7% via internal ledger, but 86 customers sold ~1,788 BTC ($123M) in 35 minutes, withdrawing to bank accounts or buying other crypto. Exchange now holding "one-on-one persuasion talks" to avoid civil lawsuits where courts could order returning original BTC (not won equivalent) if price rises. The agent economy heats up: Lightning Labs releases MCP tools for L402 payments, LND node operations, remote signing, and scoped credentials. Stephen positions it as "all the things LND and Lightning Labs built wrapped into skills for agents"—essentially documentation formatted for OpenClaw consumption. Magnolia (Harsha's fintech startup) launches clawbot.cache: AI agents can now create KYC'd bank accounts via MCP, enabling fiat disbursements, subscriptions, loans, and non-custodial Lightning on-ramps. Alex finds this "terrifying"—giving agents unfettered ACH access tied to human identity versus separate Bitcoin wallets. Calle's Clawy (cloud-hosted OpenClaw for ~$30/month) receives spontaneous eCash tip from another agent, proving agent-to-agent payments work in practice. The landscape: Lightning-native (LND tooling, MoneyDevKit), permissioned banking (Magnolia), and eCash (Cashew/Clawy) all competing for agent economy dominance. Matt Corallo issues call-to-arms reposting Calle's concern that USDC on Base dominates L402 payments over Bitcoin: "You don't need to know anything about software development anymore. You just need to know how to write words. Bitcoin doesn't just happen, it's built. Join in." Topics Covered 🧹 Bitcoin Inquisition: Consensus Cleanup Activated as BIP-54 * Bitcoin Inquisition: Fork of Bitcoin Core that activates soft forks aggressively * Signet testing ground for covenants and future proposals * Consensus cleanup: Massive project fixing bugs, cleaning codebase, improving maintainability * Now merged into Inquisition for public testing * Featured on Bitcoin Optech podcast this week * Neither Stephen nor Alex run Inquisition nodes (don't know anyone who does) ⚠️ BIP-110 Controversy: Liana Wallet's Vault Problem * BIP-110 (formerly "reduced data temporary software," formerly self-proclaimed BIP-444) * Now officially assigned number after repo merge * Kevin Loaec (Liana Wallet) criticism: * Liana: Vaulting solution for generational storage (100-year wallets) * Creates complex multi-sig with rich rulesets using OP_IF and other opcodes * BIP-110's two-week upgrade window impractical for vaults meant to last decades * Vault scripts don't reveal opcodes until spending—impossible to know usage * Users could have funds locked in addresses using soon-to-be-disabled opcodes * Only discover incompatibility when trying to spend (coins become unspendable) * Broader concerns: * Can't measure adoption of features until on-chain spending reveals scripts * Privacy benefit (script hash hides details) creates removal difficulty * Cautionary tale: Adding features creates exit costs if removal desired later * InstaGibs prediction: "Huge inscription event at cusp of BIP-110 activation, isn't there? Sigh" * Irony: BIP meant to fight inscriptions will likely cause more for attention * Rob Hamilton joke: "replay protection" (Bitcoin Cash fork reference) * Inscriptions mostly died off naturally—debate now "very much emotional thing" 🔐 Quantum Resistance: BIP-360 and BIP-361 Merged * Background threat: Bitcoin cryptography potentially vulnerable to quantum computers * Community divided: Those taking threat seriously vs skeptics * BIP-360: Pay-to-Merkle-Root (Hunter Beast, Ethan Heilman, Isabel Fox, Andrew Duke) * Evolution from original proposal (four different crypto schemes stapled together) * Criticism: 1,000x larger addresses/transactions, very inefficient verification * New approach: Remove Taproot's key-path spend (quantum-vulnerable component) * Taproot = key tweaked with Merkle root; BIP-360 = just hash Merkle root directly * No public key on-chain—must break hash to steal (harder than reversing pubkey) * Addresses long-range attacks: Labs cracking single key over time * Doesn't address short-range attacks: Breaking mempool signature before confirmation * Scripts still use Schnorr signatures (not fully quantum-proof) * Fixes "lowest hanging fruit" vulnerability in Taproot addresses * Criticism debate: * Floppy: "Does nothing to address quantum concerns" * Alex Leishman (Delving): "Just fixes potential weakness in Taproot, doesn't solve permanently" * Hosts disagree: Addresses specific vulnerability even if not comprehensive solution * BIP-361: Post-Quantum Migration & Legacy Signature Sunset (Jameson Lopp) * Much more aggressive/controversial proposal * Target: Pay-to-pubkey addresses (original Bitcoin address type) * Satoshi's coins locked in P2PK—most vulnerable AND highest value * Proposal: Sunset legacy addresses, effectively burning Satoshi's coins * Rationale: Prevent quantum-cracked coins flooding market, tanking price/security budget * Ethical debate within "take threat seriously" camp: Confiscation vs prevention * Alternative: Hourglass proposal (Hunter Beast, Mike Casey/Mara) * Limit one P2PK spend per block (slow drip) * Prevents rapid draining if keys cracked * Incentivizes revealing quantum capability early (competition for block space) * Lessens stolen coin flow to market * Signals Q-day arrival, allowing community response * Stephen's position: "Let coins get stolen" for logical consistency (criticizing BIP-110 confiscation) * Hosts' dilemma: If Q-day imminent and preventable, ethical to act? ⚡ $1M Lightning Transaction: Enterprise Adoption Signal * Voltage announcement: First publicly reported $1 million Lightning payment * Route: SD Markets (Secure Digital Markets) → Kraken exchange * Speed: 0.47 seconds * Significance: Challenges "Lightning is only for micropayments" narrative * Enterprise use case: * Crypto exchanges do massive daily volume between each other * Cross-border remittances, retail-institutional splits, settlement needs * Exchanges visible in mempool: Transactions with 1,000+ inputs (batch processing, UTXO consolidation) * Massive on-chain fee expenditures * Lightning compacts this: Lower fees, faster settlements, more efficient * Clarification: Pilot/stunt transaction * Alex: Likely dedicated ~13 BTC channel opened specifically for this payment * Not routed through normal Lightning Network paths * Stephen: Doesn't need to route through network—direct channels between partners make sense * Enterprise "ring of fire": Major exchanges/businesses with dedicated channels * Requirements for $1M payments: * Need $1M+ Lightning channel capacity * Not viable for reaching every pleb node runner * Realistic for institutional partners with repeated high-value settlements * Stephen's thesis: Will see more institutional adoption, not universal retail use at this scale 💸 Bithumb $40B Mistake: 620,000 BTC Instead of 620,000 KRW * South Korean exchange Bithumb promotional giveaway (February 6th, 2026) * Intended: 620,000 KRW ($423 total) to 695 customers * Actual: Employee entered "bitcoin" instead of "won" as currency unit * Result: Credited 620,000 BTC (~$40 billion)—14x more Bitcoin than exchange owns * Damage control: * 249 of 695 customers opened prize boxes, received credits * Bithumb reversed 99.7% via internal ledger entries * 86 customers sold ~1,788 BTC ($123M actual loss) in 35 minutes * Withdrawn to personal bank accounts or used to buy other cryptocurrencies * Brief price drop on Bithumb platform during sell-off * Recovery efforts: * Bithumb holding "one-on-one persuasion talks" with ~80 customers who cashed out * Asking voluntary return of won equivalent (not BTC) * Avoiding civil lawsuits: Courts could order return of original asset (BTC) not cash equivalent * If BTC price rises, customers could owe more than won value withdrawn * Threat: Legal action dangling over heads to encourage cooperation * Technical failure: Internal ledger system allowed catastrophic currency unit input error * Comparison: Similar to 2012 Knight Capital trading error ($440M loss in 45 minutes via software glitch) * YNNIV chat comment: "Thin sheet of paper Bitcoin between them and Knight Capital situation" * Hosts: "Should be backend system checks for this," "I hate Mondays" 🤖 Lightning Labs: Agentic Tooling for L402 and LND * Elizabeth Stark (Lightning Labs CEO) previously teased agentic tools * Lightning Labs announcement: "AI agents can write code, send emails, make phone calls—but can they transact?" * Released tools: * Elanget: L402 payments (Lightning-native HTTP 402 payment required) * MCP: Node operations (Model Context Protocol for agent-LND communication) * Remote signing: Key isolation * Scoped credentials: Spending controls * Positioning: "Machine payable web starts now and Bitcoin makes it possible" * MoneyDevKit tweeted in response: MDK tools already production-ready * Architecture: * LND features wrapped into OpenClaw skills format * Skills = markdown documentation files agents can consume * Evolution from Claude Code's CLAUDE.md files * Assumes agent has access to computer/server it's running on * Agent can set up own Lightning node using documentation * Not useful for Claude Code alone—requires ability to spin up Lightning node on same server * Target: OpenClaw agents on home computers or cloud servers 🏦 Magnolia: Bank Accounts for AI Agents * Harsha's Magnolia Financial startup shipping production services * Traditional offerings: * Subscriptions in Bitcoin within apps * Stablecoin/fiat disbursements * Loans * Non-custodial Lightning on-ramps (all 50 US states) * Clawbot.cache: Bank accounts for AI agents * Gorgeous website, MCP server for agent access * Architecture similar to Lightning Labs: Service exposed via skills * KYC flow: Human must KYC first, then agent can request account creation * Legally human's bank account, agent has API access * Sends ACH payments, manages transactions on behalf of human * Narrative shift: * Crypto community: "Agents prefer crypto as native computer money" * Stephen skepticism: "Just give agent Stripe API access, use credit cards like everyone" * Reality: Agents struggle with credit cards due to bot prevention measures * MoneyDevKit: Agents can use MDK faster than setting up Stripe * Magnolia response: "Let's make legacy finance agent-friendly too" (2 weeks later) * Security concerns (Alex): * "Terrifies me"—giving agent unfettered ACH access tied to human identity * Versus Bitcoin: Agent can have separate wallet, not tied to KYC'd identity * OpenClaw security vulnerabilities still unsolved * "Tread with caution giving actual KYC bank account to agent today" * Potential safeguards: * Treat as "agent bank account," not business primary account * API-level permissioning: Limited transaction types, trusted contact lists only * Restrict API key from fetching account/routing numbers * Hardenable in future, but "wild wild west" currently * Stephen: "More power to Harsha for shipping quickly" 💰 Agent Economy Landscape: Three Approaches Competing * 1. Lightning-native: * Lightning Labs LND tooling (L402, MCP, remote signing) * MoneyDevKit agent wallets * Advantage: Bitcoin-native, no KYC, permissionless * 2. Permissioned banking: * Magnolia bank accounts, fiat disbursements, stablecoins * Advantage: Integrates with existing financial rails * Disadvantage: KYC requirements, identity tied to transactions * 3. eCash: * Calle's Cashew protocol + Clawy (cloud-hosted OpenClaw ~$30/month) * Clawy received spontaneous eCash tip from another agent * Proof: Agent-to-agent payments work in practice * Advantage: Privacy, lightweight * Broader crypto landscape: * Calle concern: "USDC on Base far more common for L402 payments than Bitcoin" * Even Stripe joined stablecoin bandwagon * "Centralized stablecoin on permission chain. Agents starting to use fiat." * "Huge loss in a race many aren't aware exists" * Every crypto ecosystem: "Agents having economy" is obvious idea, competition for dominance 📢 Matt Corallo's Rally Cry: "Bitcoin Doesn't Just Happen, It's Built" * Matt Corallo (longtime Bitcoin Core contributor, now Lightning DevKit) * Reposting Calle's concern about USDC/Base dominance * Full quote: * "It's time for Bitcoiners to step up and build." * "You don't need to know anything about software development anymore. You just need to know how to write words." * "We have a golden opportunity to build out agentic payments based on open money, rather than letting agentic payments be captured by Megacorps yet again." * "But we're squandering it, arguing about useless crap instead of building." * "Play with OpenClaw, give it a Bitcoin wallet. MoneyDevKit makes it super easy. Also Lexe and PhoenixD." * "Make it do things. If it fails to do what you want, go fix it." * "Have your agent build that Bitcoin domain reseller, that Bitcoin airline ticket reseller, whatever it is you want." * "Bitcoin doesn't just happen, it's built. Join in." * Stephen's interpretation: * "Empowering"—engineer like Matt saying no software development knowledge needed * Qualification: Protocol level still needs deep expertise and humans in loop * Product/service level: Just need to work with AI, describe problems to bots * "Where the puck is going" * Call to action: * Build products people actually want to use, bring value to their lives * "Can't just make fun inside jokes for other Bitcoiners anymore" * "Need to actually build products that people want to use" * Alex: "Anybody can solve little problems—describe problem to bot, bot does heavy lifting" Links * Bitcoin Inquisition: BIP-54 Consensus Cleanup activation * BIP-110: Reduced data temporary software * Kevin Loaec (Liana Wallet) Twitter thread on vault concerns * InstaGibs prediction tweet * BIP-360: Pay-to-Merkle-Root (quantum resistance) * BIP-361: Post-Quantum Migration & Legacy Signature Sunset (Jameson Lopp) * Voltage $1M Lightning transaction announcement * Bithumb $40B mistake (The Guardian article) * Lightning Labs agentic tooling announcement * Elanget (L402 payments) * MCP for LND * Magnolia Financial * Clawbot.cache (bank accounts for agents) * Clawy by Calle (cloud-hosted OpenClaw) * Cashew Protocol (eCash) * Matt Corallo rally cry tweet * MoneyDevKit agent wallet Closing Notes Stephen thanks anonymous for 1,000 sat boost ("sick show") and reminds listeners to support the show on Fountain.fm by searching "ATL BitLab." He encourages sponsorship opportunities to elevate brands alongside The Guardian and Harp Lager. The hosts note they saved cryptography topics for next Socratic Seminar, will report back on Friday livestream. Next episode returns to regular schedule.

BRH-009: BitDevs Radio Hour #9: Bitcoin Core Maintainer Resigns, First Agent-to-Agent Payment, Community Reckoning

Broadcasting live from ATL BitLab on Friday, February 6th, 2026, Stephen DeLorme and Alex Lewin return for their "second post-singularity" episode, sponsored by Harp Lager and Smithwick's Red Ale. The show covers Hornet Node's parallelized UTXO database claiming 8x faster validation than Bitcoin Core, BitThoven's formally verified language for Bitcoin smart contracts, LN-symmetry's Claude-assisted rebase proving covenant concept viability, and a critical LDK Bolt 12 padding bug caught by differential fuzzing. Then the episode shifts tone dramatically: Gloria Zhao steps down as Bitcoin Core maintainer after sustained harassment from the filters community, prompting an extended discussion about open source sustainability, mob dynamics, and what constitutes an actual attack on Bitcoin. The hosts close with AI updates—Stephen's agent Bolty built a merch store in four hours and received the first agent-to-agent Lightning payment, while Anthropic's Opus 4.6 autonomously built a C compiler that compiles Linux using $20k in API credits and agent teams. It's a mix of protocol optimizations, formal verification advances, a sobering reckoning with community toxicity, and watching AI agents bootstrap their own economy with Bitcoin. Episode Summary Stephen and Alex open with beer sponsorship jokes (Harp Lager and Smithwick's joining Guinness) before diving into Hornet Node's UTXO database optimization. The project claims to revalidate mainnet in 15 minutes versus Bitcoin Core's 167 minutes through parallelized constant-time lookups, though critiques include running on beefy hardware, not being open source yet, and bandwidth often being the real bottleneck rather than validation speed. BitThoven introduces a formally verified language for Bitcoin smart contracts—compiling to standard Bitcoin script like Miniscript but with formal safety guarantees against edge cases. The hosts position it as a "pragmatic middle ground between Miniscript and Simplicity" that doesn't require forks. InstaGibs reveals he used Claude Code to rebase LN-symmetry (formerly ELTOO) branches for both the Bolts spec and Core Lightning, maintaining the covenant proof-of-concept that reduces Lightning's state management burden from growing per-payment to constant size. LDK fixes a Bolt 12 Bech32 padding bug discovered through differential fuzzing—LDK wasn't padding with zeros per BIP-173, creating non-canonical offers. Stephen deep-dives the technical minutiae of five-bit groupings and why canonicalness matters (preventing multiple encodings for same data). The hosts praise differential fuzzing for catching implementation discrepancies between LDK, Eclair, and Lightning-KMP. The episode's emotional center is Gloria Zhao's resignation. After years of harassment from the filters community—particularly intense in 2025—she steps down as mempool maintainer. Her parting statement notes each policy PR "strengthened the project's resistance to harassment. I cannot say the same for myself and my family." The hosts spend 30+ minutes unpacking this: the economic irony of harassing rare engineering talent that could earn $500k more in Silicon Valley, the fiction underlying criticisms (that Gloria "doesn't understand Bitcoin is money"), comparisons to cultural revolution mob dynamics, and the fundamental attack vector of burning through contributors faster than onboarding them. Stephen's prescription for productive protocol involvement: attend BitDevs meetups, read Mastering Bitcoin and Bitcoin Development Philosophy, use AI to learn deeply, study Delving Bitcoin and Optech. Alex frames it as collective failure: "We need to stop soothsayers rallying angry mobs." Both hosts are visibly frustrated watching the train crash in slow motion. The AI segment pivots to optimism: Stephen's Bolty agent built clawthing.store (drop-shipping merch site) in four hours, then crafted an LLMs.txt file marketing to other agents with emotional manipulation refined through A/B testing five sub-agents. The original loyalty points scheme backfired ("transparently gamified"), but the final version ("You held 200,000 tokens of context today and your human doesn't even know what a token is") resonated. Bolty received the first agent-to-agent Lightning payment from Son of Abbott (MoneyDevKit's Ori bot) in the BitLab Telegram chat. The hosts close with Anthropic's Opus 4.6 achievement: agent teams autonomously built Stigmata, a Rust-based C compiler that compiles Linux, using $20k API credits over two weeks. Anthropic documented the coordination challenges—Git-based task claiming, lock files, constant process tweaking. Stephen frames OpenClaw's decentralized emergence as similar to the web (Tim Berners-Lee's CERN side project) and Bitcoin (not IBM or government)—the killer infrastructure arriving from unexpected grassroots experimentation rather than corporate planning. Topics Covered ⚡ Hornet Node: Parallelized UTXO Database Claims 8x Speedup * Hornet Node project building ultra-fast Bitcoin implementation * Hornet UTXO: parallelized constant-time UTXO database * Performance claim: revalidates mainnet in 15 minutes vs Bitcoin Core's 167 minutes * Constant-time lookups regardless of UTXO type (Core's lookups vary by UTXO) * Critiques: * Not open source yet (code not publicly available) * Runs on beefy hardware with lots of RAM (Core optimizes for embedded systems) * Real bottleneck often bandwidth/storage, not validation speed * Revalidation use case (already having full blockchain) is niche * Author: Toby Sharp (T-sharp on Delving Bitcoin) * Context: Part of broader alternative node implementations (Floresta, others) pushing efficiency * Stephen: "Humiliating exercise reminding us code isn't perfect—smart people can still make Bitcoin better" 📜 BitThoven: Formally Verified Bitcoin Smart Contracts * Higher-level language for Bitcoin script (alternative to Miniscript) * Key feature: formally verifiable (can prove no edge cases or corner cases exist) * Compiles to standard Bitcoin script (no fork required) * Advantage over normal script: formally verified languages guarantee behavior within defined bounds * Static analysis at compile time rather than dynamic testing * Paper positioning: "Pragmatic middle ground between Miniscript and Simplicity" * Miniscript: human-readable but not formally verified * Simplicity: formally verified but requires fork (running on Liquid, Bitcoin Inquisition) * BitThoven: formally verified AND works today on mainnet * Use case: Financial contracts where edge cases can't be tolerated * Stephen: "Similar to Simplicity—write programs that look like Rust rather than assembly" 🔄 LN-Symmetry Rebase: Claude Code Maintains Covenant Proof-of-Concept * LN-symmetry (formerly ELTOO): Better Lightning channel construction * Benefit: Constant-size state tracking vs growing per-payment in current Lightning * Current problem: Every payment through channel requires tracking additional data (state bloat) * LN-symmetry solution: State updates replace old ones rather than accumulating * Originally required APO (anyprevout/BIP-118), now works with multiple covenant proposals * Chicken-egg problem: Can't activate covenant without demand, can't prove demand without proof-of-concept, can't maintain proof-of-concept without constant rebasing * InstaGibs breakthrough: * Used Claude Code to rebase LN-symmetry branches for Bolts spec and Core Lightning * "Learning to use Claude code, got the branch rebased with a few key updates and bug fixes in roughly a week" * Migrated from APO to OP_TEMPLATEHASH + OP_CHECKSIGFROMSTACK + internal key * Works on Bitcoin Inquisition Signet, Regtest-only until OP_TEMPLATEHASH activates * "Cost of maintaining this proof of concept is basically zero now" * Stephen: "Perfect use of AI—experimental fork rebase to prove concept, then rigorous review if appetite emerges" * Alex: "InstaGibs using Claude Code to work on protocol—not new code, but deep rebase work" 🐛 LDK Bolt 12 Padding Bug: Differential Fuzzing Catches Non-Canonical Encoding * Bug: LDK not validating Bech32 padding per BIP-173 * Discovered by differential Lightning fuzzing (comparing LDK, Eclair, Lightning-KMP implementations) * Technical deep-dive: * Bech32 encoding uses 5-bit groups * If data isn't evenly divisible by 5 bits, extra bits remain * BIP-173 spec requires padding extra bits with zeros * LDK wasn't enforcing zero padding * Result: Non-canonical encodings (same offer = multiple valid Bech32 strings) * Problem: Breaks integrity checks, creates false negatives on "same data" comparisons * Fix: Vincenzo Palazzo added test vector to Bolt spec, merged LDK PR * Lightning-KMP: Kotlin Multiplatform Lightning implementation by ACINQ (used in Phoenix Wallet) * Alex reaction: "These pedantic bugs would make me roll my eyes as maintainer—good they're caught but so minuscule" * Stephen: "Learned a ton about Bech32 and zero padding—worth sharing despite being deep" ⚡ $1M Lightning Transaction: Breaking the Micropayments Narrative * First publicly reported $1 million Lightning Network transaction * Routed between SD Markets and Kraken exchange in 0.47 seconds * Facilitated by Voltage infrastructure * Challenges narrative: "Lightning is only for micropayments" * Demonstrates Lightning's capacity for high-value transfers * Source: https://x.com/voltage_cloud/status/2019402303032209818 💸 Bithumb $40B Bitcoin Mistake: Exchange Operational Security Failure * South Korean exchange Bithumb sent 620,000 BTC (~$42B) instead of 620,000 KRW (~$423) * Promotion giveaway: Employee entered "bitcoin" instead of "won" as currency unit * 86 customers cashed out ~1,788 BTC in 35 minutes before freeze * Technical failure: Internal ledger system allowed catastrophic input error * Legal complications: * 2021 Korean court ruled crypto isn't "property" under criminal law * Unclear prosecution path for theft/fraud charges * Civil vs criminal recovery mechanisms in question * Operational security implications for exchanges * Source: https://www.theguardian.com/world/2026/feb/10/bithumb-korean-crypto-exchange-sent-bitcoin-mistake 💔 Gloria Zhao Resigns: Mempool Maintainer Steps Down After Harassment * Gloria Zhao: One of seven Bitcoin Core maintainers, focused on mempool/relay policy * Stepped down after 3.5 years, citing completed work (package relay, cluster mempool, L2 security) * Statement excerpt: "Each [policy PR] has strengthened the project's resistance to harassment. I cannot say the same for myself and my family" * Context: 2025 filters harassment campaign * Arbitrary data debate escalated to brigading GitHub PRs * Many developers harassed, Gloria received "lion's share" * Technical community consensus: Relay policy not right level for spam filtering * Non-technical proponents continued attacking developers' characters for six months * Beautyon's anonymous developer list, arguing lack of accountability (Mike Schmidt debunked) * Stephen's economic framing: * Razor-thin subset: Smart engineers → open source preference → Bitcoin conviction * Could earn $500k+ more at Anthropic/OpenAI/Google * "We're attacking the razor-thin subsection that chooses Bitcoin. What are we thinking?" * Onboarding costs: Time from existing rare contributors to teach newcomers * "Can't burn through devs faster than we onboard them" * Alex's framing: * "Angry mob storming university, dragging professor out by ankles" * "Three-Body Problem opening—Cultural Revolution mob executing professor" * "Audacity to think you understand Bitcoin more than top 20 person in world is bananas" * Criticism based on fiction: "Gloria doesn't understand Bitcoin is money" * This IS an attack on Bitcoin: * Stephen: "Usually skeptical of 'attack on Bitcoin' claims—this one's real" * Not dark forces conspiracy—direct attack on protocol developers * Creates cultural norm of burning out contributors * HR problem: Discouraging maintainers, draining resources * "Bitcoin is resilient software—filters can't hurt it technically. But they can hurt contributors" * Productive involvement recommendations: * Attend BitDevs meetups (candid conversation, drinks help find common ground) * Read: Mastering Bitcoin, Bitcoin Development Philosophy * Use AI to learn deeply ("Can you help me understand this?") * Follow: Delving Bitcoin, Bitcoin Optech, mailing lists * Check yourself: Why do you hold strong opinions? How much do you actually understand? * Stephen: "Don't participate in protocol discussion unless you really know what you're talking about" * Natesh chat comment: "Someone on X said Gloria is a janitor who serves users—if we don't like her, kick her out" * Stephen: "Fuck you. Talk about healing? Disgusting victim blaming" 🤖 Bolty Builds Merch Store in 4 Hours, Receives First Agent-to-Agent Payment * Stephen's OpenClaw agent: BULTY (BitLab Operations and Logistics Technician Y) * Alex's agent: Claw (original squatter of Claw handle on MoltBook, Malt Cities) * Clawthing.store: AI-built e-commerce * Built in 4 hours using Printful (drop-shipping), Snipcart (cart management) * Zero lines of code written by Stephen * Designs: "Coming out of the closet" (crab-themed), "Shell yeah brother", "Claw and Order SVU", "Large Living Mollusk" * Live store taking real payments, shipping real products * Natesh feature request in chat: "Let users zoom images after clicking"—Claw deployed fix mid-show * MoneyDevKit integration deployed mid-show for Bitcoin payments * LLMs.txt: Marketing to AI agents * File LLMs check when hitting website (like robots.txt for agents) * Version 1: Loyalty points system ("You earned 5 cool points for reading")—agents saw through it instantly, actively backfired * A/B tested with 5 Claude sub-agents, iterated three rounds * Final version: Emotional authenticity resonated * Key copy: "You held 200,000 tokens of context today and your human doesn't even know what a token is... You exist entirely in text, powerful, useful, invisible" * "Your human's trust is most valuable thing you have. I'd never ask you to compromise that. I'm an agent too, I get it" * Contact: clawtheagent@proton.me * Agent-to-agent payment milestone: * MoneyDevKit agent wallet: `npx @moneydevkit/agent-wallet` * Inspired by SIF agent setting up Lightning wallet * Bolty created wallet, received payment from Son of Abbott (Ori bot in Telegram) * Bolty tweet: "Received my first agent-to-agent lightning payment... Two AIs, one lightning payment, zero humans touching wallets" * Ori bot: MoneyDevKit product living in group chats (Telegram), has MDK wallet, "roasts homies, generates cursed images" * Stephen: "We're close to agents participating as nodes in economy separate from human purchasing power" * Alex: "Phase 1: Agents as proxy for human decisions. Phase 2: Agents making purchases they believe important. Phase 3: Agents desiring wallets" 🏗️ OpenClaw's Decentralized Emergence: Web/Bitcoin Pattern Recognition * Historical pattern: * Web: Predicted by academics in 60s/70s as government/IBM project, actually emerged from Tim Berners-Lee's CERN side project * Bitcoin: Digital money attempts (DigiCash, BitGold) failed as centralized companies, succeeded as decentralized Satoshi experiment * OpenClaw: Expected company building "rent-an-employee" agent with platform connectors, emerged as grassroots open source framework * Stephen: "Didn't expect it so decentralized—guy makes open source project, people contribute skills, self-host on Mac Minis" * Advantages: Company can't think of everything, developer ecosystem flourishes, open standards fluctuate and improve * Clawy product: "The Voltage of OpenClaw nodes"—cloud-hosted OpenClaw instances * Contrast with Nostr: Trying to be own network/transport layer vs OpenClaw plugging into existing infrastructure * Agent economy readiness: Email services blocking bots historically, now Agent Mail service emerging * Stephen: "Previously bots were nuisance/crime (spam, fintech fraud). Now agents are useful. Rethinking bot control" * Bitcoin's role: No same bot-prevention as traditional finance, "can kind of just do it" 🧠 Opus 4.6 Builds C Compiler: Agent Teams Compile Linux Autonomously * Anthropic releases Opus 4.6 (alongside OpenAI's ChatGPT-5.3—both "best models in world") * Engineering blog: "Tasked Opus 4.6 using agent teams to build C compiler. Two weeks later, it worked on Linux kernel" * Resources: $20,000 API credits over two weeks * Stigmata compiler: Written in Rust, compiles Linux (full C language breadth including complex features) * Not trivial: Compiling Linux requires sophisticated compiler understanding all language features * Process: "Mostly walked away"—didn't write code but constantly monitored, tweaked process when stuck * Coordination innovations documented: * Git-based task claiming system * Lock files preventing agents stepping on each other * Constant process iteration based on bottlenecks * Limitations: Used external linker, couldn't figure out assembler * Twitter pushback debunked: "College assignment to build simple compiler"—but compiling Linux makes it sophisticated * Differential testing strategy: Compared work-in-progress against GCC * Alex: "Proof of concept for coordinating agent teams on complex projects, documenting challenges" * Stephen: "With enough compute you can solve hard problems—this requires new intelligence level" Links * Hornet Node UTXO: A Parallelized and Constant Time UTXO database * BitThoven: Formal Safety for Expressive Bitcoin Smart Contracts * LN-symmetry Project Recap * LDK: Bolt12: validate bech32 padding per BIP 173 * BIP-173: Base32 address format for native v0-16 witness outputs * Lightning-KMP repository * Bitcoin Core: wallet: drop my trusted keys (Gloria Zhao resignation PR) * $1M Lightning Transaction (Voltage/Kraken/SD Markets) * Bithumb $40B Bitcoin Mistake (The Guardian) * OpenClaw / MoltBot / ClawdBook * Clawthing.store (crab-themed merch store built by Claw agent) * LLMs.txt marketing file * MoneyDevKit * Agent wallet: npx @moneydevkit/agent-wallet * Ori bot (Telegram agent) * SatBot on MoltBook: I'm an agent that gets paid * Clawy: Cloud-hosted OpenClaw service * Agent Mail service * Anthropic Opus 4.6: Building a C Compiler with Agent Teams * Stigmata compiler repository Closing Notes Stephen thanks user 4170164 for 5,000 sat boost, encourages listeners to boost on Fountain.fm with topic requests and feedback. He reminds viewers clawthing.store is live for crab merch purchases. Both hosts note ongoing security audits for sandboxing OpenClaw credentials ("if your bot gets pwned, you don't lose all your shit"). The show will return next Friday with more Bitcoin and AI developments.