The current WWW was built for humans - not for agents or bots. As a human, it may be helpful to see nice photos and videos to convince you to buy things. As a bot, those mediums only create friction to process information and increase costs.

But with the rise in Openclaw, Claude Cowork, and others, we think everyone will have an AI personal assistant in the next 6-24 months.

As such, we need to build an entirely new WWW for these agents to use to transact and engage on the internet. This is a 2 part call for startups -- areas that we are interested in funding at Hustle Fund:

1. Agent Identity & Authentication Infrastructure

1a) Today's auth is CAPTCHA, OAuth consent screens, and cookie-based sessions — all designed to verify humans. Agents need a portable, verifiable identity layer: "This agent is acting on behalf of EY, with these permissions, funded by this wallet, with this reputation score." A passport system for agents.

1b) Sharing of agents. Multiple people at a company may need to use the same agents. So how do you reconcile their identities and usage (and revocation of usage)?

Our portfolio company Midlyr enables auth and KYC for agents interacting with financial institutions.

Think of all of this like an Okta or 1Password for AI Agents.

2. Structured Web / Agent-Readable Content Layer

Most web content is trapped in HTML, CSS, videos, and photos for nice visual rendering. Agents waste enormous compute parsing, scraping, and interpreting pages.

What agents really need for faster scraping and processing is a parallel structured-data layer for the web. Every webpage or commerce page should have a machine-readable version in pure text or JSON.

Looking for a company to come and help make the web more readable for agents.

Our portfolio company Catalog is doing this by creating e-commerce infrastructure for agents that makes product information accessible.

3. Agent-to-Agent Marketplace & Discovery

Right now, if your agent needs a capability (translate this document, check this legal clause, generate this image), it has to know exactly which API to call. There's no "search engine" for agent services across the web.

There are people building agent marketplaces, but the bigger opportunity is to also encompass ways to learn new skills not explicitly documented in these marketplaces yet.

In addition, there should be a way to standardized capability descriptions, pricing, SLAs, and reputation.

4. Agent-Native Communication Protocol

Email is SMTP + human-readable text. In fact, Gmail does not allow bots to use its APIs (I got banned even though I didn’t send a single message).

You could create a new email system like Agentmail.to has for your agents to email. But there could be a new protocol for negotiating, coordinating, and exchanging structured information at machine speed rather than parsing through a human construct.

Our portfolio company Agent Relay is building a new communication protocol between agents.

There’s a lot in the communications layer to build.

5. Agent Compliance, Audit & Governance

When agents start acting autonomously — spending money, sending messages, signing agreements, accessing data — enterprises will need to answer: What did the agent do? Why? Was it authorized? Did it comply with regulations? There needs to be an observability and compliance layer for agent actions — logging every decision, flagging policy violations, providing audit trails, and ensuring agents operate within legal/regulatory boundaries. This is the "boring but critical" infrastructure that every regulated industry will require.

Think SOC2 sorts of stuff. This especially becomes important with the rise of skills marketplaces. How do you know they are safe?

A variation on this - perhaps this is also where the blockchain ledger comes in - should there be a transparent record of all activities agents do for public cases?

Another variation on this is agent analytics. How effective is your agent?

We’ll be back next week with more ideas that we are interested in funding at Hustle Fund.