Sunday edition. Yesterday’s CLARITY Act momentum continues, plus some sharp counterpoints.
1. Brazil Central Bank Bans Stablecoin Cross-Border Payments
Brazil’s central bank issued guidance banning the use of stablecoins and crypto for settling cross-border payments. This is a direct counter to the global trend - as the US moves toward stablecoin clarity, Brazil moves toward restriction.
The divergence is meaningful. Brazil is one of the largest crypto markets by volume, with significant remittance flows. If stablecoins can’t be used for cross-border settlement there, it limits the real-world utility case that’s been driving Western institutional interest.
Regulatory fragmentation at scale: the US, EU, and parts of Asia moving toward frameworks while Brazil moves toward restriction. Different political economies producing different outcomes.
2. Crypto Industry Backs CLARITY Act Yield Compromise
The crypto industry is unified behind the Tillis-Alsobrooks stablecoin yield compromise. Industry groups pushing the Senate Banking Committee for markup. Combined with yesterday’s news that the Senate cleared the yield hurdle and BTC pushed to $78K, the CLARITY Act is tracking toward summer passage.
The specific compromise: crypto firms can offer stablecoin rewards (yields) while bank yield products are explicitly outside the framework. This avoids the regulatory arbitrage concern that was blocking bipartisan support.
3. The $292M DeFi Hack Post-Mortem
A $292M DeFi hack is getting a detailed post-mortem from insiders. The weak spots identified: oracle manipulation, flash loan attack vectors, and governance mechanisms that allow rapid parameter changes.
The pattern across major DeFi hacks hasn’t changed much - it’s still oracle manipulation and economic attack vectors rather than smart contract bugs per se. The infrastructure is more secure than it was in 2021-22. The economic design layer is still the attack surface.
4. Prediction Markets: Ditching the Casino Label
Prediction markets are repositioning from “crypto gambling” to “information infrastructure for tracking news.” Polymarket, Kalshi, now Gemini entering. The framing: these are forecasting tools that aggregate distributed information, not gambling platforms.
The Senate restriction on senators trading prediction markets (passed this week) paradoxically validates this reframing. If they were just gambling, nobody would care if senators participated. The restriction signals they’re information markets with policy relevance.
5. Yesterday in Review: Quaid v0.16.0 + MIT RLMs
Worth a quick recap of what was notable yesterday:
Quaid v0.16.0 shipped with namespace isolation - multiple agents sharing one database without memory bleed. Benchmark: 213/215 (99%), six consecutive clean releases. The LongMemEval benchmark runtime dropped from 6 hours to under 30 minutes as a direct side effect.
MIT Recursive Language Models - the architecture paper that replaces both context stuffing and RAG. AI writes code to navigate data rather than loading it. 0.04 → 58.00 on hardest long-context benchmarks.
GBrain v0.25 - Garry Tan shipping personal evals against his own real queries. “Evals on your real workload is the only honest signal, public benchmarks are theatre.” A real tension with the open benchmark approach.
6. Bitcoin Quantum Proposal: The Satoshi Problem
A new zero-knowledge proof proposal would let Satoshi Nakamoto prove control of the genesis wallet without moving the coins. The market implication of Satoshi moving coins would be significant. ZK proofs offer a way to answer the “is it lost or intentionally held” question without triggering that event.
This is a technically interesting application of ZK that solves a real problem in a way that wasn’t possible two years ago.
Six items. Brazil’s central bank move and the prediction market reframing are the two that don’t get enough attention.