Scoop: pay-by-bank startup @Trustly has laid off about 24% of staff, 205 roles, I can exclusively report. Trustly clients include Coinbase, T-Mobile, MoneyGram: The company appears to have informed employees yesterday, with Trustly CEO Johan Tjärnberg writing in an email to employees that I've obtained, "We are not meeting the opportunity in front of us with the level of execution it demands. The way we are set up has made it harder than it should be to deliver: structures have become too complex, work is duplicated, ownership is unclear, and investment is not aligned with our priorities." While Trustly has grown revenue by loosening risk controls in order to increase approval rates in the U.S., adjusted EBITDA has shrunk as losses have grown, a Q2 presentation to bondholders shows (image below). Trustly reported adjusted EBITDA of SEK 42 million in Q2 (about USD $4.3 million), a sharp decrease from SEK 85 million ($8.65 million) in Q2 2025. The company had negative SEK 261.7 million ($26.6 million) cashflow for the first half of 2026. "Management are very focused on liquidity, with multiple levers available," the Q2 2026 bondholder presentation said. The layoffs primarily impacted the Brazil operations, with all or nearly all of that team terminated. Key legal and compliance staff in the U.S. were also impacted, sources with knowledge of the matter told me. Trustly operates in the U.S. via money transmitter licenses; a review of NMLS shows it holds such licenses in at least 41 jurisdictions. An impacted employee in Canada described the Canadian branch of Trustly as having been "dissolved." A Trustly spokesperson confirmed the layoffs, saying, "We've shared proposed organizational changes with our employees that impact around 200 roles globally. These changes are about sharpening our focus and concentrating investment behind the priorities that will help us lead the rapidly growing open banking market. We understand the impact this will have on those whose roles are affected, and we’re supporting every employee throughout the process."

Circle won by putting USDC everywhere. Now it is buying and building the places where USDC moves. CPN was built to connect banks and payment companies and route stablecoin payments across blockchains. This week Circle made it native to Arc. Eight days earlier, Circle signed a deal to buy Tazapay. If it closes, that brings 60+ banking and fintech partners and 100+ payout markets into Circle. So follow the payment. Circle issues the dollar. CPN coordinates the payment. Arc can settle it. Tazapay can connect it back into local banking and payout rails. A lot of that used to sit outside Circle. It doesn't have to anymore. Every layer Circle pulls closer is another fee, dependency or piece of control it stops leaving to someone else. Arc starts making a lot more sense from there.

Apple Pay is entering India through the part of payments UPI didn’t kill. Reuters says Apple plans to launch next month with Axis Bank credit cards first. That sounds almost backwards in a market where UPI handles 84% of digital payment volume. But cards still come with issuers, network economics and a valuable customer base. Axis alone had about 16.26 million active cards in July. Apple doesn’t need to challenge India’s biggest payment rail on day one. It can start where the old economics are still very much alive.

Robinhood is free because you're not the customer. Your trade is the product, and someone paid to see it first. Here's the mechanic. When you buy a stock, Robinhood doesn't send the order to an exchange. It sells it to a market maker, Citadel, Virtu, a handful of firms that handle most US retail flow. They pay Robinhood a cut of the spread for the right to fill your order. Selling order flow this way has at times been about half of Robinhood's revenue. The pitch was "commission-free." In a 2020 SEC settlement, the fine print о́came out: customers got worse prices than other brokers because Robinhood took a higher payment in exchange for less price improvement. $34 million in worse execution, a $65 million fine, and the model survived it. It's still core to the business. This is the oldest trick in the spread, dressed as a favor. Free-to-you means paid-by-someone, and the someone is a firm that profits from standing between you and the market. Crypto didn't invent the hidden toll booth. It just found a slower one already running on Wall Street. You never see the fee because it isn't a fee. It's a spread, and the whole point of a spread is that you don't check.

Tabby just raised $233 million at a $6.5 billion valuation. The bigger thing is what BNPL gave them access to. 25 million users. 70,000 businesses. More than $18 billion in annualised transaction volume. Most people met Tabby through Pay in 4. Now in Saudi Arabia it's already offering longer financing for some customers, up to SAR 50,000 over as many as 12 months. New licences also let it offer working capital to businesses. In the UAE, Tabby Cash is already much closer to an everyday money product: a fee-free alternative to a debit account, with a card and transfers. More than 150,000 people were using it when Tabby announced it in July. At this point I don't really see BNPL as the destination. It got Tabby into checkout. That's where it picked up both sides at once: the buyer and the merchant. From there the expansion makes sense. More credit for the consumer. Financing for the merchant. And a shot at keeping more of the consumer's everyday spending inside Tabby. Banks usually spend years trying to get into the payment relationship. Tabby started there. Now it's working its way outward into the rest of banking. That feels like a much bigger threat than BNPL itself. The bank may not lose the customer to another bank. It may lose them to the thing they first used to split a pair of shoes into four payments.

AI in payments was supposed to show up at checkout. Ant International says it is already showing up inside payment operations. The company just unveiled a broad AI-native stack across payments, accounts, FX and treasury. Under Antom, it says 89.5% of merchants have already deployed AI agents over the past 12 months, and 81.4% of payment tasks were aided by AI. That moves the automation fight somewhere more consequential than the shopping cart. Now the agent can sit inside payment operations, reconciliation, FX and treasury workflows after the sale. The buyer-side agent decides what to buy. The merchant-side agent increasingly helps decide what happens to the money next.

Arc doesn't need ARC to move money. Circle still sold $242 million worth of it before launch. That's the clever part. Gas on Arc is paid in USDC. A company can send $10 million, settle a payment or use an app without first buying some volatile network token. But behind that clean dollar experience, Circle has already created 10 billion ARC. 807.5 million were sold privately at $0.30 each. If Arc moves to Proof of Stake, ARC is supposed to handle staking, security and governance. So Circle removed the token from the customer's way. It didn't remove the token from the business model. USDC is for using the network. ARC is for owning a piece of what sits underneath it.

AI may change the buyer before it changes checkout. Mastercard is launching an agentic payment option with Alchemy this week. The agent gets a virtual card, while the user sets spending limits, purchase categories, and whether a payment needs approval first. For the merchant, the acceptance path can stay almost boring. It still receives a card credential through infrastructure it already uses. That gives agent commerce a much easier route into production. Merchants do not need to adopt some new AI payment rail before software can start buying from them. The strange part may be how little the checkout has to change while the customer on the other side changes completely.