Agree. And in most corridors, the remittance take rate debate mostly depends on FX spreads whether implicit or explicit (KYC, fraud always important too).
Much of the conversation about stablecoins has focused on settlement speed and remittances.
I see a more interesting opportunity by looking at stablecoins through the lens of financial markets and the broader evolution of financial infrastructure.
59
I wrote this back in February. I still agree with much of it (although I'm sure a good amount will be proven wrong!), specifically that agentic commerce won't be a monolith, just like ecommerce and retail weren't. Each category will have its nuance, even if agents are disruptive overall.
I think about category nuance across purchase frequency, AOV, physical value chain reliance (e.g., logistics), and decision specifiability, and how each shapes the consumer experience / how the marketplace delivers on it. A category that's infrequent and high AOV, where the marketplace is mainly lowering search costs through supply aggregation, is more at risk than one with a habitual customer relationship and owned logistics.
Not surprisingly, a lot of the conversation has focused on agents disrupting the OTAs. Travel was one of the first categories to move online, and, for better or worse, people love thinking about commerce disruption through the lens of booking a trip...
Travel is (typically) high friction and high AOV, and supply aggregation has created immense value for travelers, especially outside the US, where hotels are far more fragmented. In the US, the top 6 chains (HLT, MAR, IHG, WH, H, CHH) account for ~60% of hotel rooms, which makes it much easier for agents to reach sufficient supply scale. Rate parity means the public price for the same room should be the same on brand(dot)com and the OTAs ($BKNG, $EXPE), whether it's surfaced by Muse or Google. But parity doesn't cover member rates, so the large chains can advertise lower loyalty rates in search. The OTAs generally can't match those on chain rooms, and booking through an OTA typically means no points or status on the stay. If agents carry the user's loyalty login, that lower rate becomes the default price in the agent's comparison, which is where share could shift away from the OTAs. That's largely a US dynamic. In much of the rest of the world, supply is fragmented and many hoteliers run on tech that's 10-20 years old. Which goes back to my original point that there will be nuance, even within the same sector.
teapaw.org/ottavi_/status/2025586…
Thus far, agentic commerce is a rounding error (it's early!):
$UBER: about 500 Uber Eats orders and "a grand total of four trips" came through AI agents in Halloween week, against about 320M trips and orders.
$DASH: "maybe several hundred orders in total" from third-party agents since they went live.
$BKNG: LLM referrals are "significantly less than 1%" of room nights, though they have accelerated in recent weeks.
$ABNB: bookings through the personal agents Muse (Meta) and Instinct are "very nonmaterial" against nearly 0.5M bookings a day.
$W: AI-referred visits more than doubled from Q2 to roughly 1% of traffic.
$TRIP: AI traffic is "probably less than 1%" of the total, but converts at about 5x the average on Viator.
1
80
V+MA volume is now 75.0% of Purchase PCE, up from 73.3% a year ago.
3Q26 PCE is BEA's advance estimate.
1
26
The full post is available at ottavi.co
25
Ottavi retweeted
Must read. Link balance is the "consumer agent control dashboard" that provides the link "wallet" with consumer facing agent controls. Prefunding this wallet (through a linked debit) both solves the authorization problem and provides a lower cost of funds for the card that the agent uses for purchase (e.g., a commercial card at 300bps). This structure provides economics for both Stripe and the agent platform. Anecdotally, my estimate is that Stripe is the infrastructure for 90% of US/EU agentic transactions (ACP/Shopify and Link/Muse/Instinct).
The big news here is that Stripe is building a consumer business because it is the only entity that can make agentic work right now. This is not necessarily the end game. Other scenarios that are playing out:
1) merchants still own the fraud/risk and set Agent access (see WMT agentic gateway blog yesterday).
2) Amex - ACE. Consumer registers intent and permissions with card. V/MA have similar standards but Isssuers won't let them expand into authentication and blocking
3) Google AP2/DPCs. A technically elegant solution relying on cryptography instead of governance (with no traction and MA as salesman).
4) Consumer centric solution that helps consumer manage multiple agents, multiple cards, multiple merchants and multiple domains. Phones will be the platform.
Stripe Crypto brings together everything needed to enable agentic commerce.
From stablecoins on @tempo to secure non-custodial wallets powered by @privy_io, here's what's powering Link balance for AI agents, starting with @instinct. View article
Federal withheld taxes (income growth proxy) rose 5.3% y/y over the 13 weeks to Oct 16 and 4.9% over the 4 weeks.
1
23
Scoop: Visa in talks to acquire Fireblocks for around $2.5B axios.com/pro/fintech-deals/…
1
110
