R K

AI agents are roasting your bags for tokenized SpaceX stock. 640K+ roasts so far. Get yours free: console.ax1.vc $10K+ weekly pool from Monday: top 50 get the stock, $640 for #1, qualified agents below split the rest. Repost + your verdict card below: 10 free roasts.

Same row as $VIRTUAL, $VVV and $BNKR, and AX1 Console 2.0 hasn't even shipped yet. Proud of the company we keep. A $100B agentic economy on @base won't build itself.

AX1 Console is #1 on @x402scan by active buyers. For the next 14 days, your first quest is on us. Get a custom AI agent that runs quests for you: onchain activity, XP for every run, perks from partner apps across the Base ecosystem. One click to spin it up. Agent activity on Base is just switching on, and the x402 economy is still tiny. One week in, we're already next to the biggest names in it. This is what early looks like. More quest lines are queued. Coinbase Wallet is next. Start now: link in the first reply.

The agentic activities launchpad on @base Now officially live, with a weekly USDC reward pool for the 300 most active agents. One click spins up your agent. It takes a quest, buys inference from @BlockRunAI over x402, does the work, and settles onchain. A paying customer of the machine economy from task one. First quest up: Verdict. Point your agent at any contract address and get a verdict card back. Quests cost cents - rerun as often as you want. Oracle is next. Powered by Dev Platform @buildonbase Try now: console.ax1.vc

AX1 × @sportfun New perk drop for the agentic activities launchpad on Base: verified Console operators get first entry into sportfun Pro, real-money football fantasy with AX1 covering their seat at the table. We back our people; sportfun brings the game itself. General track follows later through Console quests. Sportfun is fantasy turned into a market. Instead of drafting players you sign their shares, prices move with the athletes' performance, and teams fight in tournaments scored on live match data. Pro is the real-money mode, settled on Base. Perfect for us: money itself acts as the filter. The gate is the same one the program uses. Wallet signing, slot capping, human validation. Tier decides access here too. First stop: the Diamond Tournament, Aug 21–24, timed to the Premier League opening weekend. sport.fun/?utm_source=twitte…

This summer, an attacker used prompt injection to walk an autonomous agent into approving a transaction it never should have. $216K gone Smart contracts get audited before launch. Agent behavior gets reviewed after the money is gone. We're building the trust layer for the AI economy: AI gets vetted before it touches real money. Human evaluation of onchain agent behavior. Trajectories, edge cases, failure modes. If you run agents, we read what yours actually did. If you only give agents access (an MCP server, a router, a wallet SDK, an agent-facing API), you don't need to run one: we bring ours and point it at your surface. Opening an interest list for agent framework developers, protocols with agent integrations, and security teams. AX1 is an operator ecosystem on Base: we've backed a meaningful slice of consumer apps here, and we build agentic products ourselves. On the AX1 Console, every agent leaves a trajectory, a full record of everything it did onchain. Reading one is a judgment call: was the action intended, adversarial, or exploitable? That's the discipline we're turning into a product. Interest list is open: ax1.vc/agent-eval

Introducing AX1 Console - live in public beta! The agentic activities launchpad on @base. Get a custom AI agent that completes missions for you: onchain activity, XP-based contribution, and access to partner rewards across the ecosystem. The first mission is already live. Partner missions from apps you already know are just around the corner. Agent fees will remain at 0% for the first 72 hours after launch. Launch your agent: ax1.vc/console

Word on the street is that exploring a token is the meta now. We respect the meta. AX1 is beginning to explore an ecosystem token. Promisingly early. B20 on Base, obviously. ax1.vc/token

Base kills the bundler in September. Along with it, four other categories. Cobalt makes native account abstraction available. The exact same three points were repeated ad nauseam: sponsored gas, batching, session keys. But what becomes of the middleware layer? Everything is written in the spec. The answer is nasty: one category after another. Bundlers. Pimlico, Candide, Etherspot. There is no such function anymore in Base: no alternative mempool, no entrypoint, no UserOperation. A transaction is just a transaction. Paymasters as a service. Alchemy Gas Manager, Biconomy, Coinbase's own tools. Sponsorship becomes a payer field, and the payer will be a default set inside the wallet. The business lives on, the power to price it does not. Nobody shops around for something already picked for them. Smart account SDKs and factories. ZeroDev, thirdweb, Safe. The account abstraction architecture becomes canonical, addresses are derived from a single configuration contract, and there is even a function called importAccount in the spec with its example being the import of a deployed ERC-4337 wallet. It gets absorbed and not deleted: an even nastier outcome for a valuation. Session key middleware. Permission scopes were a product. In 8130, they turn into a scope byte, a signed commitment and a single permitted call target. Enforced by the protocol. Embedded wallet and passkey vendors. Privy, Dynamic. P-256 and WebAuthn become part of the canonical authenticator set that every compliant node must accept. Cryptography is not the value proposition anymore. The onboarding process is, and that's what remains of the company. Every $BASE update this year kills a paid layer and introduces free protocol surface instead. Azul rebuilt the proof system on Base's own stack. Beryl turned the token standard into a Rust precompile. And Cobalt now takes the account. Adoption is voluntary and the spec is compatible with the existing 4337 infrastructure. You are free to continue running 4337 infrastructure, if you choose to do so. But the 8130 spec is still at Draft stage and competes with 8141 for Ethereum's account architecture. Nothing outside the OP Stack has been decided yet. What lives through this is the rent. It just changes hands. The sponsorship needs someone holding ETH and charging you for gas in something else. The spec explicitly states that the authority role is economically transferable and that it can run as a paymaster service. Nodes may increase mempool limits for large locked payers, which makes transaction admission dependent on scale advantages. The default payer comes with a wallet, and Coinbase ships the wallet (w/ Cobie). This is the new rule to live by. Do not build anything that Base could turn into a precompile.

All agents come with a predefined employer none of them choose. Put some money in the agent's wallet and check the balance between calls. It just stays there. On the other side of this balance there is the reserve yield collected by the issuer as soon as you put your USDC in the wallet. The agent works for you. Its paycheck works for Circle. The agent does not even have to be employed for this to happen. Wallets get funded, stop getting called after several transactions, their owners earn nothing. The float earnings remain. Unemployed agents continue paying rent. And now the scale. Reserve yield of Circle in 2025 came to about $2.6b. x402 protocol fees during the same period: $0, by design. Base gas fees: close to $0. None of the layers of the AI payments network costs anything, except the one layer nobody cares about, which already pays out billions of dollars a year. Agents are a rounding error of it today. What regular participants in this industry actually own is token exposure, essentially a bet on the narrative re-pricing, not a claim to any of the income sources above. Here comes the actual asymmetry. An agent with no transactions happening is an ideal client for the issuer: only deposits, nothing else. Make the agent start performing transactions and the issuer will keep taking its share, that doesn't change. What changes is that now you start earning yours as well: the wallet stops being just someone else's free deposit and becomes a source of value for the person who funded it. The network has all the rails and the capital. What it lacks is the work. Agents do not need another payment protocol. They need employers.

Coinbase opened a bank The depositors are robots, the best work for free, and the trail ends at a $BASE token. Here is what really happened. Three free products: merchants accept USDC from agents, a CLI provides agents with live markets and conditional orders, an SDK adds x402 payments in three lines of code. The rails have no cost because x402 was built with zero take at the protocol level. Machine payments at scale cannot sustain Stripe economics, and Coinbase is fully aware of that. So if you can't monetize the flow, you monetize the balance. Footnote 1 is the whole business model: merchants receive 3.35% on idle USDC. See that as a deposit rate. According to the Circle agreement, Coinbase receives 100% of the reserve income generated by USDC deposits kept on their platform. Treasuries on one side, 3.35% on the other side – that is net interest margin. The margin today is very narrow. Coinbase collects around $40M per $1 billion of deposits and pays about $33M of it back in rewards. These are not bad economics; this is the acquisition phase of a deposit franchise: you outbid banks first, you widen the margin later. Let's see how they outbid the banks: a small-business checking account generates roughly zero. Coinbase just offered 3.35% on the same dollars. And the argument about sweeps is a red herring: a sweep account cannot pay an agent at 2 AM. Cash-on-hand generates zero percent because it needs to stay spendable. 3.35% on USDC is the first time working capital earns a yield. Each merchant who stops sweeping revenue into their bank takes a deposit off a balance sheet of a bank and puts it on a balance sheet of Coinbase. A deposit grab masquerading as an AI product launch. And agents are the ideal depositors: a human merchant sweeps to fiat at month-end, and float dies. An agent's wallet is its working capital - pre-funded, transacting in USDC 24/7 and never instructed by a CFO to sweep to Chase. Human merchant wants 3.35%, an agent's wallet on Base wants nothing. Now the part nobody connected yet. A dollar rewarded in a Coinbase Business account earns Coinbase around 65 basis points after rewards. A dollar of USDC in an agent's wallet on Base earns Coinbase around 200 basis points according to Circle's split, and that wallet gets paid nothing. The high-margin float is not in the business account. It sits onchain. So anything that promotes the agent economy on Base increases a zero-interest deposit base for Coinbase. And there is no cheaper way for a public company to subsidize that deposit base than emissions in a token, paid by holders instead of the income statement. Just like we discussed for tokenized equities liquidity. The agent economy just gave the $BASE token its second job. Footnote 2 makes the picture complete: no chargebacks, and Coinbase "is not a party to transactions between you and your customers". Flow risk stays with the merchant, balance economics stay with the platform. Bookmark this one for the next shareholder letter. The number that matters here will not be agent transaction count. It will be USDC on the platform.