GM and happy Thursday!

Bad news is good news as BTC, ZEC and equities recover following the failed CLARITY Act cloture vote and the Fed’s first hike in 3 years. 

Good news is good news as Blockworks launches Agentic Detection in our Monitoring product, accelerating access to the insights and events that move markets. 

Below, we dive into DAWN’s USD.infra Vault launch and share insights from across the sector.

Wednesday’s price action showed a modest recovery from Tuesday’s selloff following the failed cloture vote on the CLARITY Act. The strength is notable as Wednesday threw the second punch, with the Fed voting unanimously to hike rates by 25 bps, the first rate hike since 2023. The tape is showing that bad news may not be so bad, with BTC and equity indices both closing green.

The Privacy index led on the day, up 15% and a notable outlier within the asset class. Strength here is attributable to ZEC, up 20% on the day and clearing a new 52-week high. Many remain zidelined to the privacy-focused BTC alternative, the standout winner over the past year.  

While CLARITY’s failure may have taken the air out of the room and capped the recent rally, the outlook remains constructive into year end. One component of this forecast is the term structure on Pendle’s Boros, revealing the market’s implied path of funding rates forward in time. This yield curve is now in contango, pricing in rising yields into year end. With a spread of over 60 bps between the back month and the front month, a yield curve of this shape has historically preceded both higher prices in BTC and rising onchain yields over the coming 90-120 days, as this research shows. The curve’s slope is steepest following the end of October maturity, suggesting that the upside velocity may still be several weeks out. 

When these markets move faster, so should our workflows for monitoring them. Today, Blockworks is launching Agentic Detection in Blockworks Monitoring, giving exchanges, banks, infrastructure providers and fintechs the ability to receive alerts as soon as Blockworks’ AI agents detect a development, before waiting for analyst verification. The agents continuously monitor X, GitHub, governance forums, Discord, Telegram, project blogs and more than 1,000 news sources across 1,000+ assets, classifying developments across roughly 10 categories and 55 subcategories, including hacks and exploits, outages, network upgrades, regulatory actions, token migrations and governance activity.

Every alert is labeled Agent-detected or Analyst-verified, with the latter remaining the default across Monitoring and Agentic Detection available for workflows where speed matters most. The detection layer is built on seven years of Blockworks analyst coverage, with analysts continuing to verify developments and refine the agents. The result is a workflow where agents surface developments at machine speed, while analysts remain in the loop to verify, contextualize and refine what matters. 

— Luke

Brought to you by:

Avalanche Summit NYC returns September 16-17, bringing together the institutions, enterprises, investors, and builders turning blockchain technology into real business outcomes.

From tokenized markets and institutional finance to payments and consumer applications, the Summit will explore how production-ready infrastructure is enabling faster settlement, lower costs, and entirely new products and revenue streams.

Use promo code BLOCKWORKS15 for 15% off!

The Math Behind DAWN’s Vault

DAWN opens deposits for its USD.infra Vault on Monday, and the design choices deserve more attention than the launch itself. The protocol, built on the Andrena fixed-wireless network that has connected more than 15K households across 10 US states since 2019, wants to finance the small end of AI and digital infrastructure: apartment building Wi-Fi contracts, regional ISP acquisitions, carrier offload deals with T-Mobile and AT&T, and later GPU neoclouds and training CDNs. The pitch rests on a claim we've made before: traditional project finance abandons deals below roughly $50M, and the operators locked out of that market hold signed contracts with paying customers. DAWN underwrites the contract, not the borrower or the hardware.

The mechanics follow a now-familiar template. Depositors swap USDC for USD.infra, a non-yield-bearing stablecoin issued on M0 with T-bill backing, then deposit into the vault for sUSD.infra, a participation token that accrues value through its exchange rate. Capital flows into bankruptcy-remote SPVs that hold title to equipment and contracts, and contracted revenue flows back to the vault. Participation is limited to eligible non-US persons under Reg S, with the restrictions enforced at the token level through Solana's Token Extensions rather than a legal wrapper around it, which is how a transfer-restricted asset stays composable across venues like Kamino and Exponent.

Two design choices separate DAWN from the direct lending protocols we covered in our COIL report. The first is rule-based allocation. No person or entity exercises investment discretion over depositor capital. Projects must clear binary eligibility criteria, a minimum modeled unlevered return of 20%, a 1.15x debt service coverage ratio, a servicer and independent backup servicer for the life of the contract, and documented equipment redeployment value. Eligible projects then receive a published score and get funded in score order, subject to concentration caps. Governance parameter changes take effect after a 30-day delay, so the rule cannot bend around a project already in the queue. The queue, scores, and inputs publish to a Deal Explorer for anyone to recompute. In a sector where originator quality remains the only observable proxy for underwriting discipline, a published allocation rule is a real disclosure upgrade, though it shifts the trust question: the curator verifies inputs, and an attestation provider validates deployed capital and impairments offchain.

The second choice is the liquidity sleeve, which is where the math gets interesting. DAWN plans to hold ~40-50% of TVL undeployed to support redemptions, against a floating performance target of ~12%. If around half the vault sits in T-bills earning ~4%, the deployed half must generate ~20% gross to hit the blended target. That lines up with the 20% minimum project return in the eligibility criteria, but it leaves no room for underperformance, and it makes the target a function of deployment velocity. The sector's live books show the trade in both directions. USD.AI ran a 6.4% deployment ratio at our February note and now sits at 46%, with $272.4M deployed against $586M in TVL, yet depositors still clear ~7% net against an ~8% gross because the undeployed share earns the T-bill rate. GAIB runs the opposite posture at near-full deployment, $18.7M financed against $18.9M in sAID with ~$172K in cash reserves, paying 11% net while unstaking requests queue on a monthly cadence. DAWN's 40-50% sleeve plants it closer to USD.AI's current mix than to GAIB's, and the USD.AI print is the cautionary math: a book near half deployed pays closer to 7% than 12% until the pipeline converts.

The asset mix matters too. Apartment building contracts run about 10 years with per-unit economics that do not depreciate like GPUs, and DAWN's proof-of-bandwidth telemetry gives depositors per-asset SLA and subscriber verification that voluntary disclosure regimes lack. That transparency advantage is structural, and it arrives as the category matures around it: USD.AI closed its first full repayment cycle in June and published a loan-level attestation, while GAIB's dashboard shows a consistent cadence of interest and principal payments since December. The remaining test is the one no protocol has taken: a default workout in public. Until then, DAWN's telemetry and published allocation rule are the strongest disclosure stack in the category, and the epoch-based redemption queue is the mechanism to watch if secondary prices slip below exchange rate under stress.

— Nick

Circle launches Arc mainnet with BlackRock and Visa among validators, mints 10 billion ARC tokens

Circle opened Arc's public mainnet yesterday with more than 100 applications live at launch, including an Aave v4 lending market, Morpho and Uniswap, after the testnet processed over 700M transactions since October 2025, per the company. Users pay gas in USDC with sub-second finality, and founding validators include BlackRock, DTCC, ICE, Mastercard and Visa, with DTCC planning tokenization of DTC-custodied assets on Arc starting in H2 2027. Circle also completed the genesis mint of 10B ARC tokens this week, though it framed the mint as a technical milestone rather than a commitment to a public token launch as it weighs a move from proof of authority to proof-of-stake in 2027. 

Challenging Poseidon: Pressure Testing ZK Security

LayerZero Research pressure-tests Poseidon, a widely used hash function in zero-knowledge proving systems, identifying structural weaknesses that can, under certain configurations, allow an attacker to control aspects of the hash output. The research builds on prior algebraic attacks, with the Top Gun and Slipway papers introducing techniques that reduce Poseidon’s algebraic complexity and expose weaknesses in how parameters are evaluated. While the findings do not constitute a break of Poseidon as deployed today, they suggest existing security criteria may not fully capture these attack classes and contributed to the Ethereum Foundation’s decision to move away from Poseidon for its evaluated applications.