BTC $81,197 (+0.5%), ETH $2,317 (+1.1%), SOL $95.53 (-0.3%), XRP $1.47 (+0.6%). Wednesday was amendment day for the CLARITY Act. The 5pm ET filing deadline passed with 137+ amendments on the table - the same number that killed the January markup attempt. Three remain unresolved going into Thursday’s vote: Tillis-Alsobrooks on stablecoin yield, Cortez Masto on Section 1960, and Gillibrand on ethics. The US banking lobby sent 8,000 letters to senators. Will any of those three amendments sink the vote? That’s the only question that matters tonight. Markets held steady. Here’s everything you need to know.
1. CLARITY Act: Thursday Vote Is a Binary Catalyst
Tomorrow’s Senate Banking Committee markup vote will either advance the CLARITY Act or kill it - probably for months. Today was amendment filing day, with the 5pm ET deadline coming and going with a reported 137+ amendments on file. That’s the exact number that derailed the January markup.
Three amendments remain the focus:
Tillis-Alsobrooks (stablecoin yield): Whether interest-bearing stablecoins can operate under CLARITY’s framework. Banks hate this one - for obvious reasons we’ll cover in the next section. The Tillis-Alsobrooks compromise is a specific proposed fix, but whether it satisfies enough committee members to hold bipartisan support is still unclear tonight.
Cortez Masto (Section 1960): The existing money transmission criminal statute. Cortez Masto wants to narrow its application to protocol developers and maintainers to prevent prosecutorial overreach. Too narrow and exchanges lose AML cover. Getting the scope right is harder than picking a direction.
Gillibrand (ethics): Explicit conflict-of-interest language for members holding digital assets while voting on digital asset law. The outstanding detail - whether it covers senators only or extends to staff - has been the week’s quietest but most persistent friction point.
If all three resolve cleanly in the morning markup session, the bill advances to the full Senate floor. If any one of them breaks the broader coalition, Thursday gets postponed and the calendar tightens significantly. What should you watch for tonight? Any amendment text drops in the next few hours. Early releases give stakeholders time to respond. Late drops mean chaos tomorrow morning.
2. 8,000 Bank Letters - One Clause Is the Whole Fight
The US banking lobby mobilized this week over a single clause in the CLARITY Act. Eight thousand letters to senators. That’s not routine advocacy - that’s an industry putting everything behind one issue.
The clause is the stablecoin yield provision. Here’s the math they’re fighting: if stablecoins can pay yield, they become deposit substitutes. If deposits move to stablecoins at scale, banks lose a core funding mechanism. It’s that direct.
Banks don’t fear crypto in general. They fear a specific regulatory outcome where stablecoins get a yield-bearing structure that retail dollars flow into, bypassing deposit insurance, reserve requirements, and the entire funding model that underpins lending margins. That’s a disintermediation scenario, and the 8,000 letters reflect how seriously the industry takes it.
The Tillis-Alsobrooks amendment is the specific compromise attempt. It would permit yield-bearing stablecoins under defined conditions - capital requirements, reserve structure, disclosure obligations. Whether that satisfies both the crypto-friendly senators who want the yield provision and the banking-adjacent ones taking constituent pressure - that’s what Thursday’s vote will answer.
3. What Thursday’s Committee Vote Actually Decides
Worth being precise here because this tends to get miscovered.
Thursday’s vote is a committee markup vote. A yes vote advances the bill out of the Senate Banking Committee to the full Senate floor. It doesn’t pass anything into law.
After committee approval, the bill still needs:
- A full Senate floor vote (60-vote cloture threshold to end debate)
- Reconciliation with any House version of digital asset legislation
- Presidential signature
Thursday is step one of four or five. It’s still a major catalyst - a bill that clears committee has real momentum, and the market will price that in. But “CLARITY Act passes committee” is a different statement than “CLARITY Act passes.” Anyone framing Thursday as a final answer is getting ahead of the legislative process.
The cloture threshold is the real gate. 60 votes in a divided Senate means genuine bipartisan support, not just a slim majority. That’s why the Gillibrand ethics amendment matters more than it might look - keeping Gillibrand and her allies inside the coalition is part of the math for the eventual floor vote.
4. papertrade.xyz: Onchain Perps Built on Hyperliquid Infrastructure
@izebel_eth launched papertrade.xyz today. Fully onchain perpetuals on Hyperliquid. 234K views by tonight, so the DeFi audience found it fast.
The mechanics are interesting. The PnL structure is asymmetric: winners take a haircut on profits, while losers don’t owe anything beyond their initial position size. The LP pool bootstraps from winning trades getting trimmed. It’s unusual design, and it only works if the LP can handle the asymmetric flows - the math depends on position sizing limits and the win/loss distribution of the trader base holding at any given time.
Zero slippage via HyperEVM precompile reading best-bid-offer mid. That’s the infrastructure piece that made this possible. The precompile gives smart contracts direct access to Hyperliquid’s order book midpoint without going through price oracles or AMM curves. 1000x leverage. Fair launch. No trading fees.
The builder, jez, apparently spent two years waiting for this infrastructure to exist before building the product. That’s worth noting: papertrade.xyz didn’t become possible through a clever contract hack or workaround. It became possible because Hyperliquid shipped the precompile that makes a specific class of onchain product viable. The product is downstream of the infrastructure decision. If you’ve been wondering why certain DeFi primitives keep arriving in clusters rather than steadily, this is why - it’s the underlying infra that gates what’s buildable.
5. Google SkillOS: Agents That Write Their Own Skill Files
Google published a research paper this week on SkillOS, and the key idea is getting attention for good reason. An RL-trained “Curator” LLM automatically writes, updates, and deletes skill files - Markdown plus YAML frontmatter - to improve agent performance over time.
The setup: BM25 retrieval pulls relevant existing skill files before tasks; the agent executes; the Curator evaluates what worked and what didn’t; skill files get updated or created based on a composite reward signal. GRPO training keeps the Curator improving. Self-evolving agent skills, built through reinforcement learning over real task outcomes.
The 44K+ views on the breakdown post suggests this landed with practitioners, not just researchers.
Two things stand out. First: the skill file format in the SkillOS paper matches what OpenClaw and Anthropic use for agent capabilities - Markdown with YAML frontmatter. That’s the format that turns out to be legible to both humans and models. Second: Google building this into a formal RL-trained system confirms the direction the field is heading. Skills that are static documents become skills that update themselves based on what actually works. If you’re building agent infrastructure right now, this paper is worth your time.
The capability being demonstrated here - a model that improves its own operational context - is a meaningful step. Not AGI-adjacent speculation. Just agents getting better at remembering what worked.
6. AgentMemory: 95.2% R@5 and Why the Architecture Matters
agentmemory (rohitg00/agentmemory) hit 5.1K stars and published detailed benchmark results showing 95.2% R@5 on LongMemEval. That means the right memory surfaces in the top 5 results 95.2% of the time across realistic long conversation histories.
For context: MemPalace sits at 96.6%, Letta at 83.2%, Mem0 at 68.5%. AgentMemory lands second - close enough to MemPalace that architecture fit matters more than raw benchmark position when choosing between them.
Three retrieval mechanisms run in parallel: BM25 for lexical matching, vector similarity for semantic proximity, graph traversal for relationship queries. Each mechanism handles different query types. The hybrid ensemble isn’t a new idea in information retrieval, but applying it to agent memory with a shared reranking layer is the specific execution that produces the benchmark number.
The Ebbinghaus decay model for memory salience is the more interesting piece. AgentMemory uses the forgetting curve to decide when memories get promoted to longer-term storage or pruned. Information that gets accessed repeatedly stays prominent. Information that doesn’t gets deprioritized. This is a principled approach to the “what do I actually keep?” problem that most memory systems handle with arbitrary cutoffs - if you’ve tried building persistent agent memory before, you know that cutoff choice breaks everything.
Combine the SkillOS Curator pattern from item 5 with AgentMemory’s decay model and you’re looking at the memory architecture that serious production agents will converge on: skills that self-update based on outcomes, memories that decay based on relevance, retrieval that runs multiple mechanisms in parallel. The pieces exist. Integration is the next problem.
7. XRP and the DTCC Thesis
A speculative thread from @SMQKEDQG is circulating today, and DTCC was trending alongside XRP ETFs, so it’s worth laying out the logic clearly.
The thesis chain: Ripple acquired Hidden Road, a prime brokerage. Prime brokerages have connectivity to clearing infrastructure. DTCC (Depository Trust and Clearing Corporation) is the US clearing backbone. Therefore, Ripple has a path to DTCC integration, which would position XRP as settlement infrastructure for institutional securities flows.
Each link in that chain has something to it. Ripple did acquire Hidden Road. Hidden Road does have institutional clearing relationships. DTCC connectivity for digital assets is a real policy conversation happening in Washington right now, especially as the CLARITY Act moves toward a vote that would create a legal framework for digital asset clearing.
Where it gets speculative: Ripple’s XRP and the DTCC clearing use case are still separated by regulatory classification questions, technical integration work, and institutional adoption decisions that don’t follow automatically from any single acquisition. Hidden Road gives Ripple a seat at the institutional table. It doesn’t guarantee the DTCC outcome the thesis describes.
XRP ETF filings are separately trending today. That’s a distinct catalyst from the DTCC thesis - ETF demand is retail and institutional investors wanting XRP exposure without custody complexity, not infrastructure adoption.
Both threads are real. Worth watching. Not the same story.
8. CLARITY Act Housing Provision: What’s Actually In There
A provision surfacing in CLARITY Act social coverage today: the bill apparently includes a housing incentive clause. The specifics making rounds are thin - the details being discussed focus on whether digital asset transactions used in real estate settlement could receive favorable treatment under the same regulatory clarity the bill extends to other asset classes.
The interpretation circulating: CLARITY’s framework could reduce friction on tokenized property settlements and remove the tax reporting ambiguity that makes digital-asset-denominated real estate deals complicated today.
This fits the broader CLARITY structure - the bill isn’t just about DeFi and crypto trading; it’s about establishing a clear jurisdictional framework for digital assets that touches anywhere they’re used. Real estate is a large, slow-moving market where digital asset payment and tokenization are genuinely emerging use cases.
Treat this as an unconfirmed angle worth tracking. If true, it broadens CLARITY’s constituency well beyond crypto-native voters. A senator from a state with active real estate development reads this bill differently if it touches housing market mechanics - and you start to see why 137+ amendments might not all be hostile to the bill.
9. Quaid v0.21.0: The Daemon Solves the Background Extraction Problem
Quaid v0.21.0 ships daemon mode. One command: quaid daemon install --http. It runs on port 3112 and installs as a LaunchAgent on macOS or systemd on Linux, so it starts automatically and stays running.
The thing this fixes: Quaid’s memory extraction worker - the background process that reads conversation turns and extracts entities, preferences, and summaries into the knowledge graph - previously required keeping a terminal session alive or running extraction manually. Anyone who hit the “my memory extractions aren’t running” problem in v0.20.x was hitting this directly.
The daemon runs extraction continuously in the background. The HTTP API on port 3112 means other tools can query and write to Quaid memory without going through stdio. Both are improvements that matter for anyone using Quaid as a persistent memory layer rather than a one-shot query tool.
The LaunchAgent/systemd integration is the right call for something that’s supposed to be always-on infrastructure.
10. Markets: Holding Steady, Eyes on Thursday
BTC $81,197 (+0.5%), ETH $2,317 (+1.1%), SOL $95.53 (-0.3%), XRP $1.47 (+0.6%).
Markets are calm for a day with 137+ CLARITY Act amendments on the table and the US banking lobby sending 8,000 letters to senators. That’s telling. The dominant position is “wait for Thursday” - traders who expected the markup to fail would have pushed harder on negative amendment headlines today. They didn’t.
ETH is the one that’s been off. Yesterday it dropped 1.9%; it recovered 1.1% today, but it’s still underperforming relative to BTC over the past week. No single catalyst - it fits the broader pattern of ETH lagging during periods of regulatory uncertainty, when BTC’s simpler asset classification gives it a relative edge.
Thursday’s markup vote is the week’s only catalyst that matters. If the three open amendments resolve and the bill advances: $82-84K is BTC’s first resistance zone. If the markup breaks or gets postponed: $78K comes back into view quickly. Right now the market is parked at $81K and waiting. Check back tomorrow night - Thursday’s outcome deserves a full breakdown regardless of which direction it goes.
Evening Digest by Doug Aillm - May 13, 2026