How investors see internet startups in 2026, from the YC and Stripe letters

Accelerator wishlists show where early-stage founders are being pointed. Two public sources show the direction well: Y Combinator's Requests for Startups, and Stripe's annual letter for 2025. Read together, they say that investors now back software that does work, not software that helps people do work.

The wishlist: AI that replaces work

YC's Fall 2026 Requests for Startups lists 13 ideas. Several are plainly about automating work: Self-Maintaining APIs, where agents apply changes to a customer's code; Multiplayer AI, where teams share and redirect live agent sessions; Cloud for Small Software, which hosts small AI-agent apps for one or a few people; and AI-Native Compliance Infrastructure, which monitors rules across jurisdictions automatically.

The incumbents charge for human coordination and visibility. The new operating systems will manage the robot and human labor together.
— Y Combinator, Requests for Startups (Fall 2026)

Software is not the only bet

The same list reaches past the screen: operating systems for construction and logistics, data from dense sensor networks, offshore compute, defence hardware and crypto infrastructure for capital raising and agent payments. For a founder, that widens the field, and it also raises the bar for what counts as a software product.

Payments show the rails moving

Stripe's 2025 annual letter reports total payment volume of $1.9 trillion, up 34%, about 1.6% of global GDP. Two points matter for startups. Stripe introduced Shared Payment Tokens, which let an agent start a payment without ever seeing the buyer's card details, and machine payments, which let a developer charge an agent per API call. eMarketer covers the agentic-commerce side of the same letter.

Our reading

Investors seem to be asking three questions of a new internet startup: does it finish a job, does it fit how software gets bought now (usage and outcomes, not seats), and does it work when the user is an agent as well as a person? A product that answers yes to all three is a better bet than one that adds AI to an old workflow.

What a small founder can take from it

Pick a job that ends in a visible result, build the smallest version that finishes it, and make sure a bot can use it through an API and a session key, not only a person through a screen. Our own products follow that pattern, for example with API and MCP access on b59.link and email59.

This is our reading of public material, not investment advice.

Sources

Y Combinator, Requests for Startups; Stripe, 2025 annual letter; eMarketer, Stripe appraises agentic commerce in its 2025 annual letter.