Menu #28: Beyond the Hype

This is The Menu: the weekly briefing from the Web3 Dinner Club.

What the market is saying and how the best builders and investors are making sense of it.

In this issue:

Signal, served weekly.

The dinner club is proudly sponsored by Novel Labs.

A multi-award-winning London storytelling studio building the brands of the future in AI, blockchain, and emerging technologies.

Best known for the $100m expansion to the Bored Ape Yacht Club, The Mutant Cartel World.

If you’re a startup or scale-up building a brand and looking for real go-to-market impact from those who have repeatedly built unicorns and category kings as VCs and founders... ask for an intro at the table.

A machine payment is a small payment made automatically by software rather than a person.

For example, an AI agent might pay a few cents to access data, use computing power, call an API or complete a task for another agent.

Traditional payment systems were designed around people, businesses, bank accounts, invoices and cards.

They are not especially good at millions of tiny payments made automatically, globally and around the clock.

Crypto networks, particularly when combined with stablecoins, can provide a way for software to hold money and pay for services directly.

In one line: machine payments are when software can pay for what it needs without waiting for a human to approve every transaction.

AI agents can already search, write, analyse, code and make decisions.

The next question is whether they can pay.

Not just in the sense of a person giving an AI access to a corporate card, but software holding a limited wallet, finding a service, agreeing a price and settling the transaction automatically.

That might mean paying for a dataset, a weather feed, an API call, storage, GPU time or another AI agent’s work.

It sounds futuristic, but the first signs of a machine-payment economy are already visible.

According to research from Keyrock, AI agents completed around 176 million on-chain transactions between May 2025 and April 2026, with more than $73 million settled.

The average payment was only around 31 to 48 cents, and 76% of transactions were below Visa’s $0.30 fixed-fee threshold. Around 98.6% of settlements were made in USDC.

Those numbers matter as this is not primarily about people buying expensive assets.

It is about software making very small payments very frequently.

A card is a useful tool for a person buying lunch.

It is less useful for an AI agent making 10,000 small payments a day.

Cards were designed around:

Machine commerce will look very different:

A payment of four cents is not impossible with a card. It is simply hard to make economical when fixed fees, fraud controls, account rules and reconciliation are included.

Crypto’s potential advantage is not that it is more fashionable than a card.

It is that a programmable network can allow value to move in smaller amounts, at any time, according to rules built into software.

The data also points to an early concentration risk.

If almost all agent payments are settled in USDC, then USDC is becoming the default currency for machine commerce, but the ecosystem is also becoming dependent on one issuer and one reserve model.

That is both a validation and a vulnerability.

It validates the idea that agents prefer a digital dollar for settlement. They do not want to make every API payment in a volatile asset.

But it also raises familiar questions:

Giving an AI agent a wallet is not the same as giving it a bank account.

The wallet needs rules.

It may need a maximum balance, approved counterparties, spending limits, transaction monitoring and the ability to pause or revoke permissions. The system also needs to distinguish between an agent making a legitimate payment and an agent that has been manipulated into paying the wrong party.

The hard problem is not whether an AI can sign a transaction.

It is whether we can trust it to do so within boundaries.

The most interesting effect may be the emergence of services that are too small, too frequent or too automated for traditional payment systems.

An AI agent could:

This turns the internet into something more economically active.

Today, software mostly requests information and waits for a human or company to pay. Tomorrow, software may be able to negotiate, purchase and settle by itself.

That does not mean machines become independent economic actors overnight. They will still operate under human ownership, legal rules and commercial permissions.

But the payment layer may become more granular and more automated.

The current numbers are promising, but they should not be overinterpreted.

176 million transactions and $73 million in settlement is evidence of activity, not proof that autonomous commerce has reached scale.

Some payments may be tests, incentives or protocol activity rather than durable commercial demand.

But start to look at the direction.

The first real use case for AI and crypto may not be an AI buying a house or managing a portfolio.

It may be something much smaller: a machine paying a few cents for data, compute or access, millions of times over.

That is where crypto may have a structural advantage, not because it removes all trust, but because it can make small, conditional, global payments easier to automate.

The AI economy will need a payment system built for software. Crypto may be one of the first credible candidates.

Every founder eventually discovers an uncomfortable truth: a great product is not enough.

If you cannot explain its value, earn attention and build trust, it will struggle to find customers.

Jeffrey Gitomer’s The Sales Bible is a direct, practical guide to that part of building a business.

It is less about clever closing techniques and more about the fundamentals: preparation, asking better questions, understanding what a customer actually needs and consistently proving value before asking for a sale.

Gitomer’s central idea is refreshingly simple: people do not want to be sold to; they want to make a good buying decision.

Your job is to make that decision easier.

The book’s strongest lessons for founders are:

Some of the book’s tone is very traditional sales-floor: energetic, relentless and occasionally over-the-top.

Read it for the principles, not as a script.

For the Web3 world, the relevance is obvious. There is no shortage of technical innovation, but there is often a shortage of clear communication.

Founders who can explain a complex product simply, build credibility over time and show customers why it matters will outlast those relying on jargon and hype.

W3DC: Building something valuable is step one.

Helping the right people understand why it matters is the business.

A curated, seated dinner for a small group of builders working in crypto, AI, and frontier tech.

One table. No pitches. No panels. No ego contests.

Just the kind of conversation that doesn't show up in your LinkedIn feed. The relationships that move capital, talent, and ideas in Web3 don't start at conferences.

They start at a handful of dinners with the same people, repeated over time.
Seats are limited by design.

Proudly sponsored by Novel Labs.

Web3 has spent most of its time talking about digital things.

Tokens. Wallets. Protocols. Online communities.

DePIN, short for Decentralised Physical Infrastructure Networks, takes the idea somewhere more tangible.

It uses blockchain-based incentives to coordinate physical hardware in the real world.

That hardware might include:

The basic idea is simple.

Instead of one company paying for and owning every piece of infrastructure, a network encourages many independent participants to contribute hardware. In return, they may receive payment or token rewards for providing capacity or serving users.

The blockchain acts as a coordination and reward layer.

Traditional infrastructure is expensive to build.

A company that wants to offer cloud computing may need to spend billions on data centres. A telecoms business may need to build networks. A storage provider must purchase and maintain huge amounts of hardware.

DePIN asks whether some of that capacity can be built from the edges instead.

Could thousands of people contribute storage? Could local operators provide wireless coverage? Could distributed GPU owners supply computing power to AI developers?

The attraction is not simply that it is decentralised.

It is that the network may be able to grow in smaller increments, with participants contributing capacity where demand exists.

This is where DePIN needs more scrutiny.

A token can reward someone for contributing hardware, but it cannot create customers.

A network may show rapid growth in devices, wallets or token holders while producing very little useful demand.

The key questions are:

The difference between a genuine infrastructure network and a token-funded hardware experiment is whether the economics survive after the rewards become less exciting.

The AI boom gives DePIN an obvious opportunity.

AI developers need enormous amounts of computing power, storage and data. Demand is growing faster than the traditional supply of GPUs and data-centre capacity can comfortably support.

A distributed network could potentially add supply from smaller providers, use capacity that would otherwise sit idle and create new markets for specialised computing.

But AI customers are demanding.

They care about latency, uptime, data security, predictable pricing and support. A decentralised network that cannot guarantee those things may be cheaper but still unsuitable for serious workloads.

This is where the old Web3 promise meets the old infrastructure test.

Decentralisation is valuable only if the result is competitive.

DePIN is one of the most promising areas of Web3 because it forces the industry to answer a useful question:

If the answer is yes, DePIN may become important infrastructure.

If the answer is no, it will remain another category where token activity is mistaken for product-market fit.

The winners will probably not be the networks with the most dramatic token rewards. They will be the ones that provide reliable services, attract real customers and make the token increasingly irrelevant to the user.

DePIN is not decentralisation for its own sake. It is an experiment in building physical infrastructure from the edges, and the market will decide whether that is actually better.

Hyperliquid is an unusually useful case study for Web3’s post-hype phase.

It began with a narrow proposition: build a blockchain specifically for perpetual futures. That proposition looked excessive when most networks were competing to become general-purpose platforms.

It now looks more like a strategic advantage.

The network reportedly processed $185 billion in monthly perpetuals volume in April 2026 and generated $844 million in validated trading revenue during 2025.

Its expansion into oil, gold, equities and prediction markets suggests that the ambition is no longer simply to become the largest crypto derivatives venue. It is to create an always-open, programmable market for a much wider range of exposures.

That does not make the HYPE token a guaranteed winner. Nor does it make the network immune to regulation, competition, market cycles or dilution.

But it does make Hyperliquid a useful example of what “after the hype” might mean.

The project is being judged less by its narrative and more by its throughput, revenue, liquidity and ability to attract users beyond crypto’s original trading base.

We are well past exciting technology; we are now asking whether the market continues to need what it has built.

That is the test every serious Web3 project now has to pass.

Nasdaq’s investment in Kraken parent Payward marks a significant step in the institutionalisation of tokenised equities. The partnership is designed to build an always-on market infrastructure for digital representations of stocks, supported by Nasdaq surveillance technology and targeted for launch in 2027.

But the real issue is not whether a stock can be represented by a token. It is whether that token carries the same rights, protections and claim on the underlying asset as the share it references.

That distinction will determine whether tokenised equities become a genuine improvement to capital markets, or simply a new way to sell familiar exposure.

The hype was about putting everything on-chain. The serious work is deciding what the chain actually represents.

Until next time