retail technology

The fee was never the point. The float was.

16 September 2026

Every guide to accepting crypto in 2026 leads with the same number: the fee. CoinRemitter at 0.23 per cent. NOWPayments from 0.5 to 1 per cent. Stripe at 1.5 per cent on stablecoins. Set against the 1.5 to 3.5 per cent that card processors charge, on UPay’s own figures, it reads like a bargain the retailer would be foolish to refuse. But the fee is the decoy. The real story is which token lands in your account, on which chain, and whether you can ever spend it.

What happened

The comparison sites have industrialised. The Bitcoin Foundation’s 2026 ranking lays out five gateways on fee, supported coins and settlement path, from CoinRemitter’s no-KYC crypto-only model to BitPay’s daily bank withdrawals in dollars, euros and sterling. UPay’s guide names eleven, adding enterprise infrastructure players like BVNK and CoinsPaid, the latter having processed over 29 billion dollars on its own reported figures, mostly for Europe’s iGaming operators.

The more interesting document is the one that ignores fees almost entirely. EdgeX’s 2026 stablecoin guide argues the choice between USDC, USDT, PYUSD and EURC “is less a question of market capitalization than of workflow fit.” USDC for regulated checkout and treasury. USDT where local liquidity decides whether a supplier can actually cash out. PYUSD inside PayPal’s walls. EURC for euro invoices. Same dollar peg on the label. Very different money in the hand.

Why it matters

Here is the shift a retailer has to grasp. A card payment is a single decision: accept Visa, or don’t. A crypto payment is a chain of them, and each link carries a cost the headline rate hides. There is the on-chain gas fee. The provider’s cut. The FX spread when you convert to the currency you pay rent in. The compliance screening. The reconciliation time. EdgeX puts it plainly: the real cost “includes the token, the chain, the provider, FX conversion, compliance review, reconciliation, and the off-ramp.” The 0.23 per cent was true and also almost meaningless.

Then there is the machine underneath. When Stripe re-entered this market it did not build rails. It bought Bridge, the stablecoin infrastructure company, and folded acceptance into the dashboard a merchant already knew. That is the tell. The value is migrating from the token to the orchestration layer, the software that mints, screens, converts and settles while the merchant sees only “paid.” Whoever owns that layer owns the margin, the data and the relationship. The coin is just the thing moving through the pipe.

And notice what crypto quietly removes. Chargebacks, estimated by Chargeback Gurus to have drained 33.8 billion dollars from merchants globally in 2025, vanish because blockchain settlement is irreversible. For the retailer that reads as a saving. For the shopper it reads as the disappearance of buyer protection. A card gives the customer a way to be wrong and get their money back. An irreversible payment does not. That is not a feature you advertise at checkout. It is a trust you spend.

What to watch

Watch MiCA do to Europe what it was built to do: sort the field. UPay’s guide already flags that EU businesses “must now consider MiCA licensing,” and CoinGate is being marketed on compliance rather than price. When regulation becomes the sales pitch, the low-fee, no-KYC operators do not win the enterprise account. They lose it.

The Roth Read. Stop shopping for the lowest fee. It is the cheapest number on the page because it is the least important one. Ask instead which token lands, on which chain, who holds it while it settles, and what your customer loses when the payment can never be reversed. The retailer who accepts crypto to save half a per cent, and hands a stranger’s software the float, the data and the buyer’s only recourse, has not cut a cost. They have sold the counter and kept the rent.

The agent will need to prove who it is before it can spend your money

12 August 2026

Everyone is racing to build the AI agent that shops for you. Almost no one is answering the question that decides whether it works: when a piece of software turns up at the checkout claiming to act on your behalf, who verifies that it is telling the truth?

That is the quieter half of this week’s agentic-commerce noise, and it is worth pausing on. Amid the fanfare about assistants that run errands, a smaller conversation is happening among the people building the plumbing. Writing on X this week, a GenLayer follower described the project plainly: infrastructure for “the emerging agentic economy”, where AI agents transact and interact autonomously, and candidly admitted that its “direct impact on ordinary daily routines is limited” today. That honesty is more useful than most of the hype around it. It names the gap. The agents are coming; the trust layer beneath them is not built yet.

Hold that next to the other signals crossing the desk this week. One founder put the tension exactly right: everyone wants an assistant that can run errands, nobody wants to hand a chatbot their credit card and hope for the best. Meanwhile the agents are quietly becoming the new front door to commerce, shifting shopping from search-driven browsing to agent-driven decision-making. Two facts, one problem. The demand is real. The guarantees are missing.

Here is why it matters, and it matters most through what I call the machine lens. For a generation, the retailer’s question was how to rank on the shelf, then how to rank in search. The new question is what the machine believes about you, and increasingly, whether the machine at your checkout is even the machine it claims to be. When a human shopper arrives, a brand knows roughly who it is dealing with. When an agent arrives, the retailer faces three unknowns at once: is this agent genuinely acting for the customer it names, does it have the authority to spend, and can the transaction be trusted after the fact if it goes wrong. Answer those badly and you have not built convenience. You have built the most efficient fraud channel in the history of retail.

That is the real work companies like GenLayer are circling. Not the shopping, the settling. Not the recommendation, the reconciliation. An agentic economy does not run on cleverness; it runs on verifiable trust, on some neutral way for one machine to confirm what another machine did and who stood behind it. Get that right and agents become a payment rail every retailer can accept. Get it wrong and every retailer will do what retailers always do with risk they cannot price: refuse it at the door.

This is also where the West should watch China, though not for the reason people assume. China did not win at digital payments by building better wallets. It won by embedding identity, settlement and trust inside a handful of ecosystems, so that when you paid through Alipay or WeChat, both sides knew the transaction would clear and could be resolved. The agentic economy needs the same foundation, and the open question is whether the West builds it as neutral infrastructure or lets a few platforms own the whole rail. That is not a technology choice. It is a power choice.

What to watch. Ignore the demos of agents booking dinner and buying trainers. Watch for the first serious standard that lets a retailer verify an agent’s mandate and settle a disputed agent purchase. The company that owns that verification layer will sit between every brand and every shopping agent, and take a toll on both. That is the position worth tracking, not the chatbot with the friendliest voice.

The Roth Read. Stop just asking whether an AI agent can find your product. Although that in itself is a must. Start also asking whether you can trust the one that turns up to buy it. The retailer that solves verification will accept agents as customers; the one that cannot will treat every one of them as a threat, and in a market where agents are becoming the front door, a locked door is the same as a closed shop.

China did not spend last week building video tools. It was building eyes.

5 August 2026

Every retailer who watched three Chinese labs ship video models in seven days filed the news under marketing. Cheaper ads, faster content, a problem for the agency. Wrong drawer. What landed last week was the perception layer for the machines that will one day walk your shop floor, and it landed open, and it landed cheap.

The launches came within days of each other. ByteDance’s Seedance 2.5 now generates thirty seconds of video with sound in a single pass, holding characters, scenes and camera logic together across a whole narrative rather than one lucky shot. MiniMax’s H3 does fifteen seconds with stereo audio generated jointly rather than bolted on, and on 3 August MiniMax put the weights on Hugging Face for anyone to download. Alibaba closed the week with Qwen3.8-Max, 2.4 trillion parameters, which now sits second in the world on the public leaderboard for reading images and visual material.

So run the telescope the other way. A model that keeps thirty seconds coherent, objects that persist behind an obstacle, weight that falls the way weight falls, a cup that is still on the table after the camera moves, has not learned to draw. It has learned how the world behaves. Generation is only the exam. The syllabus is physics, permanence and consequence. And the same weights that let a machine imagine a scene let it read one.

Reading a scene, fast, in bad light, with a person moving through it, is the entire job of a robot’s eyes.

That is the triangulation, and it is why this is a retail story rather than a media one. China already holds the other two legs. By industry counts it ships the overwhelming majority of the world’s humanoid robots, and TrendForce expects Chinese output to nearly double this year, with Unitree and AgiBot taking around eighty per cent of shipments. It holds the motors, the batteries, the rare earths, and the appetite to put machines in public before the ethics committee has finished its report. What it lacked was sight worth putting behind the visor. It is now building that in the open and giving it away.

Price finishes the argument. DeepSeek’s V4-Flash update, the least photogenic of the week’s launches and probably the most consequential, costs roughly three cents to run the full Artificial Analysis intelligence battery, against $3.15 for the Western frontier. A robot has a battery, not a data centre. Perception has to be almost free before it can live inside a body on a shop floor, and last week it became almost free.

A caution worth keeping. None of this means the machine understands anything. Video models still get physics wrong in ways that are amusing in a clip and unacceptable in a machine holding a bottle near a customer’s child. A convincing picture of a grasp is not a grasp, and that distance is where the next two years of the argument will be fought.

The direction, though, is not ambiguous. The robot that eventually watches your shelves, greets your customer and judges whether that customer is confused or annoyed will not be running perception you bought from a vendor you can name. It will be running weights someone downloaded, most likely trained in China, tuned by a supplier three tiers below the name above your door. And it will still be your brand doing the looking.

What to watch. Not the next video demo, and not the next leaderboard. Watch for the moment a Chinese maker ships a robot whose perception layer is one of these video models, and says so. When those two industries start sharing a checkpoint, the cost of machine sight collapses the way the cost of machine text already has, and every question retailers assumed they had until 2030 arrives early.

The Roth Read. Stop treating AI video as a marketing line item and start asking who supplies your machines’ eyes, because you are about to buy vision the way you buy electricity: from someone else, invisibly, with no say in how it was made. Put that question on your risk register this quarter. The machine watching your customer speaks for you, whoever trained it.