Hey folks,

This week Pete and I were joined by designer Skyler Westby, who's been shifting his focus from design to building AI agents through his new company, Finite Computer.

We talked about whether software as we know it is becoming less permanent, with agents increasingly able to create applications and interfaces on demand, rather than forcing people to work within existing software.

We also explored why owning your AI agents could become just as important as owning your business, the growing importance of human judgement as AI takes on more of the work, and why Skyler thinks the future might involve spending considerably less time in front of computers and more time with other people.

Most software is built to be permanent. You open an application, navigate its interface and work within the features someone decided to build. Skyler Westby thinks AI agents could change that. Rather than forcing people into a fixed application, an agent could create an interface for a particular task, adapt it as needed and potentially discard it when the work is done.

Email is a good example. Skyler finds himself constantly navigating messages designed to capture his attention, each pulling him towards another website, interface or action. An agent could sit between him and his inbox, filtering out the distractions and delivering what actually matters. Even a simple chat interface might be preferable to opening Gmail and dealing with everything competing for attention.

That raises an interesting question about the future of design. If agents can create interfaces whenever they're needed, who decides what those interfaces should look like? Skyler thinks there's still a role for designers in determining what people see and how information is presented. Over time, agents might learn which types of interfaces work best for particular tasks and simply generate them when needed.

Alex Gladstein at the Human Rights Foundation had been using AI agents to produce board presentations, annual reports and other documents from information shared by his Freedom Tech team. After seeing one of the finished PDFs, designer Skyler Westby suggested building a simple tool that would let colleagues leave comments directly on the document and have an agent make the requested changes.

About 30 minutes later, Gladstein had a working application. The PDF was displayed on a large canvas where colleagues could leave comments on specific parts of the document. An AI agent could then pick up those comments and make the changes without anyone having to manually work through the feedback.

Gladstein started experimenting with the same approach for video, allowing people to comment on individual elements and have an agent revise them. Skyler estimated that the process saved around two weeks of editing work. He also saw the potential for organisations to create similar tools whenever teams needed to review and revise material together, rather than relying on conventional editing software.

Imagine a family business that's been operating for fifty years. Its owner has spent a lifetime building it and intends to pass it down to their children. Now imagine that much of the work once handled by employees is being done by AI agents, using software the business doesn't actually own.

That was the concern Pete Winn raised during the discussion. As businesses increasingly rely on agents to manage their day-to-day operations, the question of who owns those agents becomes important. If they're controlled by an outside provider, the business could find itself dependent on technology it has little say over, including how it operates or what it costs.

The business might still own its premises, equipment and customer relationships, but if someone else controls the software doing much of the work, how much independence does the owner really have? For Pete, ensuring businesses own their software and agents is becoming an important part of preserving what they've built.

If you found even one useful idea in this week’s Intelligence Snacks, we have a small favour to ask. Please forward this newsletter to someone who you think would get something from it too.

Sharing it is one of the simplest ways to support Intelligence Snacks and helps these ideas reach people we might never reach ourselves.

You can find more Snacks from this week’s podcast at intelligencesnacks.com.

See you next week!

Cheers,
Pete & Andy