Building in Crypto: The Best Time Is When No One Is Looking — YC's bet on stablecoins, agent payments, and the bear-market builder

This is the tenth paper in this series reading Y Combinator's Fall 2026 Requests for Startups one request at a time, taking each one seriously enough to ask what it would actually take to build. After defense, small-software clouds, multiplayer AI, and operating systems for the physical world, the list arrives somewhere less fashionable: crypto, in the middle of a down market, argued for by Nemil Dalal under the deliberately contrarian title "The Best Time to Build in Crypto."

What the request actually says

Dalal's argument is not a price call. It is a timing call built on three planks. First, that the boring, durable parts of crypto — stablecoins and payment rails — kept compounding while the speculative parts deflated: stablecoins are being adopted by major financial institutions, fintech companies like Deel and Gusto are building on crypto rails, and regulatory frameworks have clarified. Second, that software agents are the next customer for these rails: in the request's words, "it feels inevitable that agents are going to use crypto networks as financial rails." Third, the cultural claim in the title: "bear markets let the real projects build and thrive while bull markets are the worst time to build," because downturns clear out the yield-chasers and leave the field to people who want to ship.

The request is specific about where YC wants applications: capital-raising infrastructure, new stablecoin applications, agentic commerce, trading platforms, institutional products, and scalable private blockchains. It cites funded companies as proof of the pattern — BlindPay and Infinia building financial infrastructure for Latin America, Aspora on remittances to India — and points to Hyperliquid as evidence that a small team can now challenge incumbent exchanges. YC has backed over a hundred crypto startups and expects the number to grow; Dalal's sharpest line is the prediction that eventually "every YC startup [will] use crypto rails from capital raising to payments, though most will probably never even know." YC is acting on its own thesis: starting with the Spring 2026 batch, any funded startup can take its investment in USDC.

Why now, concretely

The "regulatory clarity" plank is the most verifiable. The GENIUS Act, signed into law on July 18, 2025, gave the United States its first federal framework for payment stablecoins — defining who may issue one, how it must be backed, and which regulator supervises it. Whatever one thinks of the details, it ended a decade in which the legal status of a dollar-pegged token was a matter of enforcement-action archaeology. Institutions that could not touch stablecoins before can now build compliance programs around a statute, and implementation rulemaking has been under way since late 2025.

The agent plank is younger but no longer hypothetical. Coinbase introduced x402 in May 2025 — a payment standard built on the HTTP 402 status code that lets software pay for API calls in stablecoins without cards or accounts — and later donated the specification to the Linux Foundation. Google, with Coinbase and dozens of partners, launched the Agent Payments Protocol (AP2) in September 2025 to handle the harder half of the problem: proving an agent had authority to spend, and deciding who is liable when it spends wrongly. Industry reporting through 2026 describes tens of thousands of active agents and rapidly growing transaction counts on these rails. The absolute dollar volumes remain small; the direction is not ambiguous. When a buyer has no hands and no credit history, a bearer instrument with programmatic settlement and no chargebacks is a natural fit — which is precisely why the authorization problem, not the settlement problem, is where the risk concentrates.

The bear-market plank has a real track record, with a real caveat. Coinbase was founded in 2012 after crypto's first collapse; Ethereum was built through the 2014–15 winter; Uniswap shipped into the depths of 2018. The pattern is genuine: the companies that define each cycle were built in the trough of the previous one. But the argument carries survivorship bias — bear markets also killed thousands of equally sincere projects — and it quietly assumes the builder can survive years of thin demand. "Fewer tourists" is a hiring and focus advantage, not a revenue plan.

What is actually hard

The hard parts of this request are mostly not blockchain engineering. Stablecoin payments companies live or die on the fiat edges: banking relationships, licensing in every corridor they touch, fraud when settlement is final and irreversible, and treasury management across chains. The GENIUS Act creates clarity by creating a licensing regime — clarity is not exemption, and the compliance burden now has a statute's precision. Capital raising is harder still, because it is a securities problem wearing a payments costume; tokenizing an instrument does not tokenize away disclosure obligations, investor caps, or transfer restrictions, and the startups that pretend otherwise become enforcement actions in the next cycle. And agentic commerce inherits an unsolved trust question: a human's card comes with a chargeback right and a liability framework refined over fifty years; an agent's stablecoin payment comes with neither, which is why AP2-style mandates and human approval gates are being designed in from the start rather than bolted on.

The field attempting this is unusually credible for a bear market. Stripe bought Bridge to own stablecoin infrastructure and now runs x402 payments on Base; Circle went public on the strength of USDC; Visa is settling in stablecoins across multiple chains. Beneath the giants sit the YC-shaped companies the request names — BlindPay, Infinia, Aspora — doing the unglamorous corridor-by-corridor work. A builder answering this request needs a wedge those players will not chase: one corridor, one asset class, one agent-payment niche, executed with the compliance handled transparently and the blockchain invisible. That, more than any protocol choice, is what the request is actually asking for.

Where Gwen stands

Precision matters about where Gwen sits relative to this request: adjacent, and not more. Gwen has no stablecoin, wallet, or crypto integration of any kind today. What it does have is a working, if conventional, version of the problem this request describes. Gwen's customers buy a Work Budget; its own billing and metering rails enforce spend ceilings; and every outward action that moves money passes through a human approval gate inside a long-lived mission with a durable transcript. That is the mandate-and-authorization architecture that AP2 is trying to standardize for agents generally — Gwen simply settles in ordinary currency on ordinary rails. If agent-native payment standards mature, the work-budget model is the part of Gwen's design that points toward them, because a budget with enforced ceilings and approval gates is exactly the object such protocols need on the buyer's side. The capital-raising half of the request is familiar territory too, at one remove: a sister venture, Backers, operates a Canadian capital-raising platform where investor money moves by bank transfer under exempt-market rules — a reminder that the constraints in that business are regulatory before they are technical. Gwen builds and hosts web applications from plain-language descriptions, runs marketing and operations work, and routes tasks across many AI models by measured cost and quality. None of that requires a blockchain yet. If Dalal's prediction holds — that most startups will one day use crypto rails without knowing it — the honest version of Gwen's position is that it would be one of the ones who never knows, and that would be fine.