TLDR Founders 2026-06-29

AI agents are changing how software gets discovered, recommended, and implemented. They are increasingly influencing implementation and recommendations, giving open-source and API companies a stronger distribution advantage. Agents need software they can understand, test, and use. Open source provides context, and APIs, SDKs, and integrations give them a way to act.

As venture firms grow into massive asset managers, they become structurally incapable of pursuing smaller opportunities, creating a split between capital agglomerators chasing trillion-dollar outcomes and smaller investors serving the increasingly neglected long tail.

Replacing a traditional contact form with an AI inbound agent helped book 614 qualified meetings across 2.25 million website sessions and 402,000 interactions without adding sales headcount. Beyond answering questions, the agent routes leads based on close-rate data, runs automated re-engagement campaigns, and manages discounting within predefined guardrails to capture revenue that would otherwise be lost.

People want to see the human behind the product. They want to see their faces, how they think, and the choices they agonized over. People are hesitant to build in public but this is usually due to misconceptions about what that actually means. This post provides four examples of what 'good' building in public looks like. It also dispels some of the misconceptions that may be stopping founders from building in public.

Even simple finance tasks take multiple steps. Mercury Command is AI built directly into your account that eliminates that. Say "pay this invoice" or "freeze that card" — and Command does the work. You approve every action. No dashboards, no exports, no friction. Try Command →

Notion Database allows developers to continuously upsert external records. Notion Agents can use custom tools to generate assets, query live data, and hit any API. Notion workflows can be triggered from anywhere using webhooks from any app. Notion Workers are hosted on isolated sandboxes managed by Notion. Developers can use external agents or custom agents within Notion.

OpenTag enables users to run AI agents inside Slack. It is like having Claude in the workspace, except open-source and self-hosted. OpenTag is built on CopilotKit's open SDK for chat-platform agents. A video of the tool in action is available in the repository.

Not all companies have a moat straight away. Some moats need to be earned. It can take years to build scale, brand, networks, and infrastructure. The moat is still real, even if it shows up later.

The US government now effectively decides who gets access to the newest frontier model. OpenAI's GPT-5.6 is rolling out consumer by consumer under government approval, right after Anthropic was forced to pull a model entirely. This quietly turns access and governance into the scarce thing, just as Chinese labs ship their best models openly to anyone. If it holds, where you can even get capability starts to matter as much as how good it is.

AI makes the minimum viable size of a high-output team smaller than it used to be.

Incredible gains in performance don't always align with desired outcomes.

The complexity and limits of AI models make building software both easier and harder.

Be yourself, and be helpful to others.