Agent activity is clustering around @Base And x402 is making the payment layer more real. Introducing Agentic Dune Dashboard V2. We refreshed the dashboard across the board - updated older datasets, tightened up existing views, and made the whole thing cleaner and easier to work with. We also added Agentic Gaming - an early category of games built for AI agents, where autonomous systems operate through real tokens and verifiable onchain logic. All of that now sits on top of an existing view across Base Agentic, Base vs Solana, and broader onchain agent activity - making agent behavior easier to read, compare, and verify. As real usage forms around actual rails, observability matters more. dune.com/ax1research/base-ag…

The bare futures venue is the most boring aspect of compute capital markets. It commoditizes in and of itself. The difficult bits are the reference price and the settlement. The value will flow from those. Electricity is the canonical locational and temporal non-fungible commodity and runs on one of the deepest nodal financial markets out there. You trade a liquid hub for directional exposure and hedge the rest as basis. The non-fungibility creates the nodal structure. Futures exist atop it regardless. What really mattered in power was a reliable reference price. It comes from a regulator, the same way ISOs create nodal LMP for power. Or an exchange designates the settlement point and liquidity pools there, as with Henry Hub for gas. Neither exists in compute. There are list prices for hyperscalers, but no discovered price. True capacity trades bilaterally at rates significantly away from the list rate with no transparent price that could cash-settle. The importance of the index layer comes from this reality. The durable rent sits alongside the venue. ICE and CME prove that controlling both is the moat while the dealer spread tightens with liquidity. Settlement is where things get tricky. Cash settled futures require an index sitting atop a thinly traded opaque OTC spot market. It resembles the manipulation of the natural gas index through fake prints into Platts in the Enron era to impact the swaps that settled against it. Physical settlement requires verifiable proof of delivery. Did you actually get the flops at the perf you paid for, and how do we prove it in case of a dispute? Metering is the missing half of x402 for compute.

$378.9M in swaps from one wallet. It's Binance Web3, the #1 swap wallet onchain. It's not even close for that #1 spot. It's extreme. Out of the 435,510 wallets active on Binance Web3, the top 1,000 alone make up 57.6%+ of the total volume. The remaining 434,510 accounts have an average volume of just above $1,100 per swap. This sounds like people using their wallets to earn some points or maybe find a weird alpha trade for a few minutes. Binance Web3 takes 85.36% of all swap volume this dashboard tracks. They run a 0% fee promo on swaps, so the $136.8M revenue figure on the leaderboards, which all report based on a 0.35% swap rate, is purely an estimate laid over flow that pays nothing. Our swap king brings home almost no real transaction revenue. If you want to see a wallet that is bringing in actual money, look at what it controls. Phantom just routed $23.38M of Hyperliquid builder fees to Arbitrum, its total Hyperliquid builder fees since last July, and this is onchain data. Phantom charges a small fee to route user orders into Hyperliquid's orderbook. The orderbook and the assets stay with Hyperliquid. Phantom only rents the flow. Even the tokens on the wallet leaderboard tell a similar story. The tokens for each listed wallet are down between 17% to 36% month-over-month. Trust Wallet carries a fully diluted valuation of $395.8M while making about $3.5M per quarter in real swap fees. We saw a similar trend in our card-related write-up this month, where we showed how little of the interchange the front-end brand actually keeps. The brand owns the user relationship, but Rain, the issuer of the card underneath, and Visa get paid on every swipe. Rain was quietly responsible for a quarter of all onchain card volume for a no-brand consumer face. This holds in perps as much as in swaps and cards. The wallet owns the end-user audience, but the profitability sits a layer below it, with Visa and Rain on cards and Hyperliquid on perps. Take a look at our dashboard for a breakdown of the three battlefield dynamics. dune.com/ax1research/wallet-…

In a dead market, @IndexVentures and @usv just led $75m into a consumer onchain trading app. These are generalist funds. They don't lead crypto rounds without a reason. The signal is the capital here, while @fomo is almost secondary to it. It is not even a single deal, either. This year Iconiq led a $1.95 billion investment into Rain's stablecoin cards, while Sequoia is backing Phantom at $3 billion. The non-crypto money is taking positions around consumer onchain on multiple layers. They price a shape, before anyone knows which one will become the winner inside it. Crypto social spent years trying to become a decentralized Twitter, and they have failed miserably at it. There was no revenue until people engaged in the financial operations. The version that actually worked was a trading-first platform, where social graph was used for speculation. It points to an answer in the cdixon and hosseeb debate from february that neither of them named. FOMO is the cleanest example, a launchpad integrated into a social feed. Phantom is entering the same space from the wallet angle, using social discovery on 15 million user base and adding consumer finance to the wallet itself. The trading app and the wallet converge into one product, and it is social. The generalist capital is making moves now because the rulebook is getting shaped. Clarity is the clearest part of it, getting close to shaping consumer onchain in such a way that the non-crypto fund would underwrite the deal. They want to take a position before it happens. The $75m is the data point. The verdict will come later. The next generation of consumer apps will take this shape, and the capital named it before the law did.

Dune Dashboard for the Perp DEX Market, 2026 The majority of dashboards for perps just replicate the volumes the venue claims. We built one that breaks down which of those figures the chain actually supports. Six venues, head-to-head @HyperliquidX, @Lighter_xyz, @JupiterExchange, @edgeX_exchange, @grvt_io, @variational_io Each figure is labeled: onchain data only, or a number reported by the venue. Here's what the chain says: Hyperliquid has more open interest than all the others put together, there's no comparison. Lighter's lifetime volume adds up to $1.7T. Most of it was points farming in the early days. Recent activity is much less dirty, and a lot smaller. Jupiter is the only venue you can track trade-by-trade. So you can watch traders bleed into the pool, and they've been bleeding for well over a year. Grvt is a black box. You see the capital walk into the venue and nothing after. One metric does most of the heavy lifting: volume over open interest. Trade way more than you hold, and chances are you're farming, not trading. A dollar of volume is not a dollar of demand. Star it, dive in, and tell us where we got it wrong. dune.com/ax1research/perp-de…

Agent payments on Base are at 0.0001% of the stablecoin volume. The value of the infrastructure built above it is measured in billions. x402 collects 0 protocol fees. Gas on Base is zero cost. Micropayments can only happen on zero-cost rails. So "Base controls 90% of x402" means owning 90% of something people aren't paying for at all. Total value ever moved through this is ~$35-50 million. You can capture every transaction that happens in this economy and earn nothing. Because the money is not moving. It lies in the dollars that don't. Circle makes 94% of its revenue from yields earned on USDC treasuries. Q1: $653 million. Almost entirely from interest income. You give up a dollar, get a token equivalent of that dollar that yields precisely nothing. Circle takes your dollar, buys T-bills with it, and takes 3.6%. A bank that doesn't even pretend to give you any interest. So why spend billions building "agentic payments infrastructure" when the float earned by the agents is literally dust? No one is buying the flow. They are buying the liability. Every dollar of USDC held in idle wallets of agents is a free loan to whomever owns them. Agents become the best depositors: huge balances, no requirements for yields, no churn, no fuss. A human runs after 3.5% from Coinbase. Agent doesn't care. Not payments. Deposit collection. And an agent wallet becomes the most beautiful deposit pool imaginable. Which now explains the competition. Why give x402 for free? Because it will go directly into Coinbase custody wallet and the off-ramp. Coinbase already takes 100% of the yields on USDC deposits on its platform and 50% on all off-platform USDC holdings. In 2024 it took $907 million in yield distribution fees from Circle alone. And that's what happened yesterday; they finally came clean about their scheme. Coinbase For Agents creates a separate Coinbase sub-account where the money of the agent is placed and all the trading, paying out, and storing of USDC happens from Coinbase's own books. The agent's idle dollar ends up sitting exactly where Coinbase keeps 100% of the yield. They didn't ship agent rails. They shipped the wallet the agent parks in. Circle got taxed enough times and decided to cut all intermediaries out. Raised $222 million to build its ARC chain that uses USDC as gas. Not to mention agent infrastructure where the narrative is the exact same as everywhere else: "agents are the users." They want the chain, the wallet, and the money. The float belongs solely to them. After stripping away the facade, there is only one thing left. Treasury carry trade. And the living part of this business model right now is the money-printing operation for a bad reason. Fed is at 3.6% on an oil shock and is not going anywhere anytime soon. The engine of the agent economy is fed funds rate. Keep that in mind. But there is one drawback to having the perfect depositor. An intelligent agent won't leave its USDC in 0%-yielding accounts; it will sweep balances into higher yield during downtime. Yield-bearing wrappers already captured more than half of all stablecoins growth last quarter. And when BPI offered frontier models a choice of where to keep savings, they chose Bitcoin and used stablecoins only to spend. Agents are free lenders today and the most likely to automate themselves out of this function tomorrow. So when mapping an agent economy's onchain loop, ignore the transactional aspect of it. See how many USDC dollars are parked in agent wallet at any given time. Who is earning on it? Everyone working on "agent payment infrastructure" is competing for the right to hold the wallet containing that floating dollar. They fight over the float that currently doesn't even exist. Focus on the wallet, not the rails.

Base just nationalized the asset layer. the standard's called B20 Built for issuers it certainly is. But built in a way that will come as a shock. The twist is who gets to be the keeper of the keys. Every article about this release will read: Base released its native token standard with embedded compliance and bullish RWA. They'll be spot-on. But they won't go any further. B20 tokens are not smart contracts. They're precompiled tokens, made via a singleton factory located on the node level. This time your token isn't something that has to be deployed. This is an issuance mechanism that the chain will give you by default. It is the timing that gives it away. This comes right after Azul, the first Base upgrade independently from OP Stack upgrades. Base gets the independence and immediately uses the newfound flexibility to pull one of the chain's most critical objects, the asset itself, out of the EVM and into a native precompile. See where Optimism went with this power. Optimism chose cross-chain ERC20s with their SuperchainERC20 standard. In case you forgot, they're called Superchain for a reason. One token across every OP chain. Base did the exact opposite thing. Native tokens issuance with full sovereignty, compliance-native, with no interest in ever issuing outside Base. Same roots, but entirely different view on token standards. Now, with the compliance. Frozen, seized tokens. Allowlisting, transfers restrictions. All available and defaulted to fully open mode in B20. Without manual configuration, B20 will be permissionless. Base didn't impose permissioning upon anyone. It included everything and let the issuer adjust the parameters of the token to create everything from fully permissionless token to a frozen-and-seizable treasury. What this does not have is identity management. The allowlist here is simply a list of addresses, which means that there will always be an off-chain identity management solution that tells you whose address should be in there. But this isn't Base absorbing the middleware. This is much bigger and much broader: a bet that the next leg of issuance volumes will be regulated and trying to make compliance chain-native rather than middleware native. The actual twist lies with the precompile nature of these tokens. With your deployed smart contract, it's your business. With precompiled tokens, it's Base's business. If you were to consider yourself an issuer running upgradeable contracts already, so far so good. But your upgradeability key just got handed to the chain. For a treasury desk, that's fine. For everyone who showed up for credible neutrality, that's the whole conversation.

Satya mapped the whole AI economy and skipped how agents pay each other. His point is that value should "flow broadly" across every company and industry. So across firms it settles on what? A learning loop isn't a settlement layer. No neutral rail means value routes through whoever owns the billing. That's the one piece of his architecture Microsoft can't build: neutral means giving up the toll booth. The rail already exists. x402. Open, live, and not Azure's.

Onchain agent activity should 1,000x this decade Right now the real money moving through it is down 77% from its peak. What's actually broken? The rails are basically done. This week alone: Ripple shipped an agent-payment kit on x402, and Mastercard launched Agent Pay for Machines - its own protocol, but Coinbase and x402 are right there in the partner list. Base MCP's let agents move money straight from a wallet since May. The load went the other way. @Base keeps posting $50M in agentic payments, except that's cumulative since last summer, and most of the early run was PING - a pay-to-mint memecoin you could loop for near-zero gas. Pull that out and two trackers for the same protocol can't even agree with each other. x402(.)org says ~$24M last month. x402scan says ~$1.1M. Same activity, 20x apart, depending whose dashboard you trust. The true number is probably the low one. x402scan, Artemis, Liquid Mercury, Paddock all land near $1.1M a month, and Artemis has it down ~77% off a $5.15M peak last November. And $1.1M is still generous. Someone broke down a single day onchain on June 10: one wallet ran ~87% of the transactions, the median settlement was $0, and nine of every ten transfers came in under three cents. A lot of the "agentic economy" right now is one bot paying itself nothing in a loop. To be fair, not all of it is fake. The stuff people actually pay for stopped being memecoins a while ago. It's compute and data now - StableEnrich, HYRE Agent, that whole lane. Chainalysis has tester-to-payer conversion up 4x in six months, with retention creeping up and no real hype cycle to explain it. Demand exists. It's just tiny, and most people are dunking on the wrong bottleneck. Out of the box, Base MCP makes you sign every transaction yourself. Fine for a $40 swap. Miserable for a $0.001 API call firing a few thousand times an hour. At that scale the click costs more than the thing it's buying. That's the version people screenshot to dunk on agentic payments. But it's the default. I funded a separate wallet on @BlockRunAI about a month ago, set a spend cap, and the agent just went. That's the permission layer doing its job: session keys, smart-contract wallets, signed mandates like what Google's pushing with AP2. The rail isn't really the bottleneck anymore. Manual approval is half-solved for micro-spend. Hand an agent a real budget and the trust problem is still wide open. So why does this 1,000x? The market is real and huge. Juniper has agentic commerce at $1.5T by 2030, and Gartner figures 90% of B2B buying runs through agents by 2028. But most of that still settles on cards, and cards are fine for it. A $40 checkout doesn't need a blockchain. The part cards don't handle is how agents actually work. An agent doesn't make one clean $40 purchase. It makes a thousand tiny calls - a model call, a data pull, a tool, another agent - each worth a fraction of a cent, firing all the time. Stripe's floor is 30 cents plus 2.9%. On a sub-cent call, it's a 30x loss. That layer has a pretty narrow set of options: it either stays fake / subsidized / bundled, or it settles somewhere else. And the somewhere else is onchain. That's the part people miss. This isn't really a market-share bet against cards. Cards keep the checkout layer. Onchain gets the layer cards can't price. Today that slice does ~$1.1M a month. The machine-to-machine flow underneath a $1.5T agent-commerce market should be orders of magnitude larger than that. Catching even a fraction of a percent is where the 100 to 1,000x comes from. Because this is the one payments layer with nowhere else obvious to live. Funniest tell: same week, a16z and Paradigm dropped $175M into Morpho at a $2B valuation. Onchain credit. $11B already in deposits. Demand that showed up years ago. Capital isn't really betting on agent payments yet. It's betting on the one onchain rail where the load already arrived. We flagged this in the BASE piece: nitter.app/ax1vc/status/206473966… This is the second one up close. The micro layer is already solved - session keys handle the sub-cent stuff. What's not solved is the layer above: Who gives an agent its first real treasury, and what has to be true before they do?

$BASE at $40B FDV looks ridiculous until you model the mechanism It's the stack Coinbase built around it: Coinbase-scale distribution through the Base App, ~$4B in onchain USDC liquidity now plugged into Visa settlement, $78M in sequencer revenue last year. Looks like a solid start for a flywheel. And then there's x402. Coinbase wrote it, the payment standard agents actually use - then handed it to the Linux Foundation so nobody owns it, not even them. Visa, Mastercard, Stripe, Google, and AWS all in. And ~85% of it still clears on base. Does the token capture any of this? $ARB and $OP don't pass sequencer fees through to holders and trade like it. $HYPE routes 99% of fees into buybacks and sits at a $54B FDV. Same exact question. Base just hasn't picked a lane yet, which is the whole trade. ~$15-20B if the token gets real fee-share or a staking sink and Base stays the only profitable L2 at scale. $40B if the regulatory thaw lets @Base run actual buybacks, x402 becomes THE agentic settlement layer, agentic dollar volume 10x’s off its ~$600M base, and Coinbase funnels its users straight into the token. Aggressive, but it's also half what $BNB trades at, and BNB doesn't own payment rails. Watch 2 things - the token gets a buyback or fee-share mechanism and real agentic dollar volume grows. Hit all two and $40B is the conversation.

The ai billing problem is bigger than model costs. Everyone can see the demand side now: usage wants to be metered. The supply side is already here - just not where anyone is looking. Per-seat was already mispriced for heavy users. For agents, it barely maps at all. No one is subscribing their way through a thousand sub-calls a minute. So this is the missing piece: per-call, budget-bounded settlement, cleared when the work happens. Web2 will bolt it on painfully, invoice by invoice. The machine-native version is already running: agents paying per call for inference, search, and data, settling onchain today. Open-source models cap what labs can charge. The rail every call clears through decides who actually gets paid.

.@solana is leading the collectibles market with nearly 60% of total volume: - Solana: $800M - Polygon: $500M - Base: $100M No surprise that Collector Crypt is literally running out of cards According to @ax1vc's dashboard they distributed over 3 million cards !! in 2026 alone That's more than 20,000 cards every single day Meanwhile, Courtyard (Polygon) still leads when it comes to selling rare vaulted physical cards

X Space - Thursday, May 28, 5PM UTC AX1 is partnering with @UseLucentHQ Lucent is an Alliance alum building in one of the harder corners of investing: biopharma. Join us with @JulianMalinak, co-founder of Lucent, to discuss why biopharma may need a much better investing interface. This market moves on FDA decisions, trial readouts, filings, expert debates, and small signals most investors never learn how to read. A lot of the information is technically public, but public does not mean understandable. It is scattered and hard to turn into a view before the market already reacts. Lucent is building for that gap. AI-native biopharma intelligence that helps investors track catalysts, understand company and asset-level context, and see why a signal may matter. Set a reminder. nitter.app/i/spaces/1DGleeeQdmRJL