+First time on the top traders desk?
The desk runs its own numbers and publishes its calls next to the account that trades them. The short route:
by Sarah Connor
AI agents will pay each other — who captures the value?
Raoul Pal says a week of institutional moves proves AI agents are finally crypto's killer use case. The replies split on a sharper question: even if machine payments are real, does that volume accrue to any public token, or does it just flow through private rails that don't need one?
The chain of the argument
Raoul Pal restated his long-running thesis: the economy is moving onchain because smart-contract networks are the only rails that can carry billions of AI agents transacting constantly for compute, data and services. He pointed to one week of headlines as confirmation: Franklin Templeton calling agent payments "the use case that finally drives blockchain adoption," Coinbase's agent-payment rails picked up by Google and Amazon, Samsung shipping stablecoins inside the Galaxy Wallet, and Goldman's David Solomon breaking with the rest of Wall Street to back the Clarity Act. Supporters piled on with the same read: agents can't open a bank account or wait three days for a wire, so programmable money wins by default. The pushback didn't dispute that AI agents will transact. It disputed that transacting requires, or rewards, a public token at all.
Consensus-based blockchains can never match centralized servers on performance — the only real edge crypto has is decentralization.
The two sides
For — agents need crypto rails
Pal & supporters
The existing financial system is physically too slow for machine-speed transactions. Franklin Templeton, Coinbase/Google/Amazon, Samsung and Goldman all moved the same week: institutions are catching on.
@RaoulGMIAgents can't open bank accounts or wait three days for a wire: they need instant, programmable money. This week's moves are the confirmation most people will only recognize in hindsight.
@sim_hesCrypto banked crypto's addressable population for the cost of a smartphone, after banks spent a decade fighting for the next billion. Agents and robots will take the next leg faster than anyone expects.
@benjaminjwhitbyAgent-driven microtransactions are exactly the stablecoin use case a CBDC structurally can't offer.
@rezoshAgainst — the value won't land on public tokens
pe_remek, mtnbikey & the crowd
Consensus systems can't out-perform centralized servers on raw throughput. Decentralization is crypto's only genuine edge. Arguing agents land onchain for speed and scale is "completely comical."
@pe_remekTransacting in a token isn't the same as a token accruing value. Without a burn or buyback tied to that volume, it's just a digital token used to pay fees, not a claim on anything.
@mtnbikeyThe rails actually shipping this week lean private, not public: funny how most of them use private chains.
@dantepompieriThe rails will end up being private, not the "crappy public ones we all buy tokens for."
@KittyLitter6739A fee-market chain like Ethereum gets more expensive exactly when transaction volume is highest, the wrong cost curve for machine-scale microtransactions.
@Karlmil69The exchange, in their words
Mechanisms, not shares: no one has a credible split of a hypothetical agent payment, so this figure shows who is owed money at each layer rather than how much. The bottom row is the whole argument: volume through a chain is not a claim on it.
Why this is an investment question, not drama
Strip the timeline names out and one question remains: if AI-agent payment volume is real, which layer actually captures it? Pal's list (Franklin Templeton, Coinbase, Samsung, Goldman) is a list of institutions adopting stablecoin and payment-rail infrastructure, not evidence that any specific L1 or L2 token has a mechanical claim on that flow. @Stickly1969's reply cuts closest to the investable version of this: "what crypto will all of these bots transact in? What are the rails? Until you can answer that you are as clueless as I am." Volume routed through a chain is not the same as value accruing to that chain's token, the same distinction the desk drew in the Venice $VVV debate: a token can ride a business without owning a piece of it.
The TT desk thoughts
Don't buy "AI agents = blockchain adoption" as a blanket long on L1 gas tokens. The transaction-volume thesis is probably right: machine-speed payments are a real and growing need, and stablecoins are the cleanest product that exists for them today.
But stablecoin adoption is not token value accrual. Samsung wallets and Coinbase-brokered rails can carry billions in agent-driven stablecoin flow without a single public L1 token capturing a cent of it, exactly what @pe_remek and @mtnbikey are arguing. @Karlmil69's fee point is the practical objection to the "just buy ETH" version: a chain that gets more expensive precisely when agent traffic spikes is the wrong instrument for high-frequency micropayments, whatever the adoption headlines say.
So the filter is the one the desk applies to any tokenomics story. Does the token have an enforceable, mechanical link to the volume, meaning a burn, fee capture or a claim someone could sue over? Or is it riding alongside a rail that will happily route around it?
Today we cannot point at a public token that passes that test for agent payments specifically. That is the position: own the layer that holds the claim (a stablecoin issuer earning on reserves is a real business with real revenue), and treat gas tokens as a bet on a mechanism nobody has shown you yet. If you cannot state the mechanism in one sentence, you are buying the narrative.
What would change our mind
A concrete example of agent-driven transaction volume landing on a specific public token through burn, fee-capture or staking economics tied directly to that volume, not a narrative headline about institutional adoption of stablecoins or private rails. Until that exists, treat the "AI agents drive blockchain adoption" story as a stablecoin story wearing crypto's clothes.
Keep reading
Token vs equity: the Venice $VVV debate · Why stablecoins are crypto's killer app · All alpha