[Headline]

Welcome back to Founder Mode!

This episode brings back two repeat guests, Viktor Petersson of Sbomify and Screenly, and Alex Finn of Henry Intelligent Machines, who sit on almost opposite ends of the same question: how much should you actually trust your AI agents?

🎧 Listen to Founder Mode Episode 77 here:

Cage or Release Your Agents?... Oct 8 · Founder Mode 40:48   ​

​

Viktor runs every one of his agents through a web proxy that gives it just enough access to do its job and nothing more. He still hasn't given an AI agent full access to his email, even though he eventually let a human assistant in. Alex went the other direction. He held out on email access for over a year, then connected everything once he saw the productivity gains, and now runs what he calls an ambient AI setup, with local models constantly watching his codebase for issues in the background.

Both of them are also building real companies on top of these tools right now. Viktor's team of three or four moves faster than a company many times that size would have a few years ago, in part because he's stayed bootstrapped on purpose, worried that raising money locks him into a roadmap before he's ready. Alex did raise capital for Henry, but says it was never about the money. It was about the access and distribution his investors bring, since anything he ships will get copied within minutes anyway.

I closed the episode with five takeaways, straight from the show:

1. Decide where you stand on access first. How much are you actually willing to give your agents before you build anything else.
2. Hand off the work you don't like first. Nothing is more motivating than offloading the tasks you dread.
3. Switching tools is cheaper than you think. Ask your current agent to brief the next one, and it will.
4. Reviewing what your agents produce becomes the new bottleneck. The job quietly turns into babysitting.
5. If you raise money today, raise it for the people and the distribution, not to build a bigger team.

🎧 Listen to Pretty Good AI Episode 7 here:

Conductor Of Complex Organiz... Oct 6 · Pretty Good AI Podcast 24:43   ​

​

My guest this episode is Gordon Sampson, VP of IT Infrastructure and Operations at Planned Parenthood Federation of America, about three months into the role when we recorded. Before this, Gordon spent years in university healthcare IT, including Georgetown's medical center, so he's spent his whole career running technology for systems he doesn't fully own.

Gordon's approach to walking into a new organization starts with the most basic question possible: who's your customer? Not the org chart, the actual functional teams and the people they serve. Only once that's settled does he start layering in data, tools, and a real source of truth. He describes his job as a kind of federated model, where individual sites keep their autonomy, but everything still needs to roll up into something he can actually govern and report on.

On AI specifically, Gordon is blunt that the conversation always has to start with the data. Clean, secure, centralized data has to come before any AI tool can succeed, the same lesson he says he learned twenty-five years ago when organizations were dumping unstructured information into early machine learning systems and getting exactly the mess you'd expect. He also pointed out a strange shift happening right now: the tech industry, which spent decades trying to move faster than government, is now the one asking government to step in and slow things down with real guardrails.

Here's how I summed up the conversation:

Final Thoughts

Both conversations this week come back to the same question: who's actually in control, and what are you willing to hand over to get speed. Viktor and Alex are answering that question for their own inboxes and bank accounts. Gordon is answering it for an entire federated healthcare system that spans dozens of independent sites. The answer in both cases starts in the same place: know exactly what you're protecting before you decide how much access to give away.

Watch the new Founder Mode episode, then listen to episode seven of the Pretty Good AI Podcast. Subscribe so you don't miss what's next.

This podcast builds on the Founder Mode newsletter.

Let’s build.

-kevin

​​​

2810 N Church St #87205, Wilmington, DE 19802
​Unsubscribe · Preferences​