ChainFeeds’ Aug. 12 digest pulled together five major research threads
ChainFeeds published a research digest on Aug. 12 that grouped five areas of discussion into one briefing: Hermes’ rise in personal agent products, Andreessen Horowitz’s take on crypto payment cards, Vitalik Buterin’s updated view of Ethereum’s roadmap, a broader retreat by corporate crypto buyers, and the fee structure behind meme token launch and trading platforms.
Hermes was described as a product-layer winner, not a clear infrastructure leap
In the IOSG Ventures section, the central question was how Hermes can coexist with Claude Code and Codex through differentiation rather than direct confrontation. The digest said the agent sector already had relatively mature infrastructure before OpenClaw appeared, but adoption was centered on development projects and enterprise workflows instead of individual users. Early frameworks were built for developers, usually producing code or configuration. They established the groundwork for agents, but did not truly deliver the agent itself as a finished product.
That left the sector with a basic limitation: engineering thresholds stayed high, agents remained stuck in the phase of developer tools, and the product loop needed to turn the technology into a personal asset for end users was missing. The digest argued that OpenClaw’s real shift was not a reinvention of the underlying agent loop or task scheduling technology. The change came at the product layer. In its framing, LangChain answers how to build an agent, while OpenClaw answers how to have one.
By packaging fragmented framework capabilities into a complete product that an individual can configure and keep using over time, OpenClaw moved the unit of adoption from a project to a person. The digest said that shift showed up in several ways: giving agents persistent names and identities, using high-frequency communication apps such as Telegram and WhatsApp as the access point, and turning the agent into an always-on system rather than a passive tool waiting for calls.
Hermes, the digest said, does not appear to hold a major generational edge over OpenClaw at the level of underlying packaging capabilities. Its breakout growth was instead tied to a different fit with the market. It picked up a user base that OpenClaw had already educated but that was still struggling with operations and maintenance pain points. Hermes’ core product assumption was to shift operational responsibility away from the user, removing much of the cost tied to setup, debugging and ongoing maintenance.
The report framed that around “delegation trust.” Users need to be able to hand over an objective to an agent without constantly stepping in to repair the process. It broke that trust into three parts. One is reliability trust, supported by continued task progression, failure recovery, persistent Kanban and a /goal mode. Another is security trust, supported by approval flows, sandboxed environments and strict permission boundaries. The third is verifiable trust, supported by mechanisms such as Completion Contract and Grounded Citations so users can check whether a task was actually completed.
The digest also said Hermes’ MIT open-source strategy creates a structural commercial challenge even as it helps ecosystem growth. A free self-hosted option means the paid version has to deliver enough irreplaceable value, or users can simply deploy it themselves. A deeper risk lies in value capture: if Hermes continues to be integrated by cloud providers as an optional runtime, it could end up in a position similar to Linux or Kubernetes, influential in the ecosystem but with a large share of commercial value captured by hosting and compute platforms instead.
That leaves Nous with a key question, according to the digest: how to convert a stronger ecosystem position into direct business returns. Hermes is currently aimed at four dense power-user groups — self-hosting and infrastructure users, multi-model arbitrage users, multi-agent coordinators, and open-source and crypto AI communities. The note ended by saying future competition in personal agents may center less on one-off execution and more on task ownership and accumulated trust.
a16z said crypto payment cards are now processing more than $750 million a month
The digest’s second featured item summarized an English thread from a16z on crypto payment cards. The firm said those cards have grown from an early novelty into a spending tool with monthly transaction volume above $750 million. They let users pay with crypto at merchants that accept traditional card networks. In practice, stablecoins — which now make up the large majority of usage — are typically converted into local fiat at the point of sale, so the transaction looks little different from a standard card payment on the merchant side.
These cards do not necessarily require a traditional bank account. Depending on product design, users may deposit stablecoins with the issuer or hold on-chain assets in self-custody and spend through the card. The digest said this is expanding access to dollar-denominated assets and dollar payments for users globally while also giving stablecoin holders a simpler way to spend.
In July 2026, monthly crypto payment card volume reached $759 million, up about 2.5x from $306 million a year earlier. When tracking began in October 2023, the figure was still below $1 million. The digest noted that the data reflects on-chain activity tracked by Paymentscan. For RedotPay, the largest project by transaction volume, spending figures were disclosed by the issuer rather than observed entirely on-chain.
Usage count has climbed in parallel. Users made nearly 9 million crypto payment card transactions in July 2026, versus about 5.2 million a year earlier. Based on those figures, the average ticket size now stands at about $86.
The settlement-chain picture has changed as well. Early payment card activity was concentrated on Gnosis because Gnosis Pay, the first Visa card product directly linked to a self-custody wallet, was built there. As more card projects launched, settlement spread across more blockchains. By July, Optimism accounted for about 29% of crypto payment card volume, while Solana and Base each handled around 19%. Gnosis, which had once dominated, had fallen to about 2%.
The digest treated that as a move from a single-ecosystem setup toward a multichain payments base. As more card products enter the market, more chains are taking on stablecoin spending and settlement functions, widening the route from on-chain dollars into everyday spending.
The stablecoin mix has shifted just as sharply. Early crypto card spending was driven mainly by euro stablecoins. In early 2024, about 88% of card volume was settled in EURe, much of it on Gnosis. By July 2026, EURe’s share had dropped to roughly 2%. Dollar-based stablecoins had taken over. USDC now handles about 58% of crypto payment card volume, and USDT about 26%. A year earlier those shares were around 48% and 7%, respectively.
The digest said crypto card spending is now heavily concentrated in digital dollars. The market remains small compared with the trillions of dollars processed each month by traditional card networks, but the growth trend is strengthening. For the payment card projects currently being tracked, the expansion has happened almost entirely on top of Visa’s existing payments infrastructure.
Vitalik’s update pointed to privacy, post-quantum readiness and formal verification
The third featured item summarized an English thread from Vitalik Buterin. He said he had updated the 2023 roadmap diagram and overlaid its items with their positions in the current Strawmap. There is still substantial overlap between the two, but the center of gravity in Ethereum research has shifted in visible ways.
According to the digest, some priorities have been moved upward, with quantum security cited as one example. Other older directions have lost ground, including VDFs and many lines of EVM improvement. Some older proposals have also been replaced by better technical paths. Verkle has gradually evolved into Unified BT and then into PBT. State expiry, meanwhile, has given way to new state types.
The biggest change may be the appearance of entirely new directions that were not part of the 2023 roadmap. The digest said Ethereum is placing much more weight on strong privacy, naming keyed nonces, recent roots, parts of FOCIL, lean privacy pool and wormholes among the relevant ideas.
It also said Ethereum has started to focus on scaling for a post-quantum era, including LeanSPHINCS signatures and aggregation as well as zkzk frames. These designs are not only about performance. They are also about preserving network security in a future environment where quantum computing matters.
Another notable shift is protocol simplification, described as lean-ification, to make more complete formal verification possible in the future. The digest said full formal verification of the entire Ethereum protocol had long been close to unrealistic, but modern AI tools are making that objective more reachable. Formal verification is presented as a way to prove protocol correctness and improve infrastructure security.
Blob and gas futures also appeared in the current Strawmap as new ideas. Native rollups were listed as another new direction. The digest said SNARK technology was not mature enough in 2023 to support that design, which is why it was absent from the earlier roadmap. With zero-knowledge proving systems advancing quickly, native rollups are now emerging as an important area in future scaling research.
At the execution layer, Ethereum is also opening room for a broader redesign. zkzk frames suggest the protocol may eventually expose some non-EVM instruction set architecture to users. The main candidates named in the digest are LeanISA and RISC-V. Those ISAs are described as simpler, more modern and more efficient than the EVM. If Ethereum eventually has such a base execution architecture, letting developers target those ISAs directly could make more sense than keeping the EVM as the protocol’s central fixed execution environment. In that case, the EVM could shift into the role of an intermediate representation rather than remain the hard core of the base protocol.
New state types point to another basic change in how Ethereum may scale. The digest stressed that this is not merely a replacement for state expiry. It reflects a broader rethink: instead of trying to scale every form of Ethereum activity as far as possible, the protocol could adopt more specialized mechanisms tailored to specific needs and constraints. That idea is visible in both new state types and zkzk frames.
One thread running through several of these directions is the use of STARKs and AI-accelerated formal verification as core elements of Ethereum’s future system design. The digest said recursive STARKs could appear across multiple protocol layers, and that a special primitive called “aggregate to union verified dependencies” is expected to be used in the execution layer, consensus layer and data layer.
Buterin’s broad conclusion, as summarized by the digest, is that Ethereum is moving into a more mature phase. The network is aiming for quantum security and stronger user privacy while preserving security and censorship resistance, then pushing for higher performance and broader scalability on top of those base goals.
Corporate Bitcoin buyers were portrayed as the next group under pressure
In its market commentary section, the digest cited TechFlow’s argument that enterprise buyers and institutional participants are now facing a more visible squeeze. Strategy has sold Bitcoin for two straight weeks. In the latest week, it sold 1,690 BTC at about $64,262 per coin, raising $108.6 million, with all proceeds used to repurchase STRC preferred shares. The digest said Strategy’s average Bitcoin holding cost had previously been about $75,385. Based on a current price near $63,900, its roughly 840,000 BTC position was down about 15% overall.
Trump Media was cited as another example. According to the digest, when Bitcoin was near its all-time high in the summer of 2025, Trump Media bought about 9,500 BTC at an average price around $108,519, spending more than $1.1 billion. Less than a year later, the fair value of that holding had fallen to $557.1 million, leaving a paper gap close to $500 million, while quarterly revenue was only $1.7 million.
At the same time, Trump Media management scrapped a plan with Crypto.com to jointly list a CRO treasury company. The company’s CEO said on an earnings call that it would “pull back some crypto and online entertainment expansion” and refocus on social media. In the digest’s framing, crypto assets in such companies are shifting from a strategic expansion line into a side business where risk control and loss containment take priority.
The note placed that development in a longer cycle. It argued that crypto differs from traditional industries because it depends on periodic and intense clearing events to improve market structure. The 2018 ICO bust removed large numbers of projects and speculative funds that lacked real value, while leaving infrastructure such as Ethereum and DeFi. The 2022 collapses of FTX and Luna cleared out highly leveraged lending models and opaque centralized exchanges, pushed the sector toward greater on-chain transparency and compliance, and indirectly helped create the environment for spot Bitcoin ETF approval in 2024.
The digest said the 2026 clearing phase is now reaching corporate buyers. Strategy’s loss-making sale shows that even a long-term Bitcoin holder must react when its capital structure is under pressure. Trump Media shows that corporate treasury allocation still has to follow investment discipline or it becomes a balance-sheet problem during market drawdowns.
It also cited Grayscale’s ETF lineup adjustment as a sign that institutional attitudes toward altcoins are changing. For tokens beyond Bitcoin and Ethereum, ETF narrative alone is not enough to win acceptance in traditional finance. Real demand, liquidity and long-term value are taking on more weight. The digest added that mNAV falling below 1 means the earlier cycle of issuing stock to buy crypto, then using rising crypto prices to support the stock and raise more capital for further purchases, is breaking down. Cash flow, balance-sheet quality and capital efficiency are coming back into focus.
Meme issuance and trading platforms are still producing large fee streams
The final major section summarized an Odaily breakdown of revenue models across meme-related platforms. It said the broader meme sector has cooled from prior bull-market levels, but remains one of the few areas still showing persistent activity. From CASHCAT, which gained traction on Robinhood Chain, to the stock-themed meme token MarsCoin on BNB Chain, new hotspots are still appearing and still pulling in community attention.
Pump.fun, the largest meme token launch platform, generated $34.68 million in revenue over the past 30 days even in a weaker market, with about $1.718 billion in trading volume over the same period. The digest said its earning power has even surpassed Hyperliquid. The main revenue source is continued trading while newly issued tokens remain in the bonding curve phase. Token creation itself is free, but users pay transaction fees when buying and selling in the bonding curve. The current total fee rate is 1.25%, with 0.95% going to the protocol and 0.30% to the token creator. When a token graduates from Pump.fun into PumpSwap, an additional 0.015 SOL graduation fee applies.
Flap and Pons were highlighted as other fast-growing token launch platforms. Flap generated about $5.58 million in revenue over the past 30 days, with more than 90% coming from BNB Chain at roughly $5.05 million. Robinhood Chain contributed about $529,000. Platform trading volume during the period was about $908 million. Flap also uses a bonding curve launch model, then migrates tokens to a DEX after the curve stage. It allows creators to issue tax tokens with self-defined trading tax rates, and MarsCoin was issued through Flap. The base protocol transaction fee is 1%.
Pons, native to Robinhood Chain, had at one stage become the largest meme token launch platform in that ecosystem. In mid-July, daily token issuance exceeded 15,000, and the platform held the top market share for roughly half a month. As of Aug. 11, Pons had generated about $18.76 million in fees over the previous 30 days, of which protocol revenue was about $4.99 million. In Pons V1, creating a token cost 0.0005 ETH and trading fees were 1%, with around 70% going to the token creator and about 30% to the protocol. After the upgrade to V2, the platform’s revenue base expanded to include launch fees, bonding curve trading fees and ongoing fees after tokens graduate to Uniswap V4. The team also plans to use 80% of protocol revenue to buy back and burn PONS.
On the trading-tool side, GMGN, Axiom and Fomo were described as fee-driven businesses. As of Aug. 11, GMGN had posted $2.584 billion in 30-day trading volume and generated about $23.87 million in fees. After invitation rebates, protocol net revenue was about $19.81 million. Robinhood Chain was the biggest revenue source at about $11.67 million, followed by BNB Chain at about $4.93 million and Solana at about $2.41 million. GMGN’s business model is straightforward: the platform typically charges a 1% fee on each trade and returns part of that through referral rebates.
Axiom remains centered on Solana. Over the last 30 days it recorded about $1.337 billion in volume and about $23.91 million in fees. After rebates and trading cashback, protocol revenue came to about $14.67 million, almost all of it from Solana. Fomo, which has drawn recent attention, leans more heavily into social trading. Users can view other traders’ live positions and rankings and can copy trades directly. As of Aug. 11, Fomo had generated about $8.79 million in protocol revenue over 30 days, with around $8.64 million from Solana spot trading and about $153,000 from Hyperliquid perpetuals. Under its pricing rules, spot trades carry a minimum fee of 0.5% of notional and at least $0.95 per trade, while perpetuals add a 0.05% platform fee.
Taken together, the digest sketched a market where personal agents, stablecoin payment rails and Ethereum base-layer research continue to move forward, even as corporate Bitcoin allocation and parts of the institutional narrative lose momentum and meme-linked platforms keep pulling in substantial fee income.
