Staying current with AI pays off when you can test a new model against real work and get it into a live trial over a weekend. That's what I did with Jev in a client's email workflow, at about a quarter of the cost per call.

As AI moves into every app, users are being asked to pay again for intelligence they already buy elsewhere. My move from Raycast to Tinycast shows how that changes software pricing, adoption, and moats.

Better attribution gave me an answer I didn't want. Without an agent checking the web server logs, I wouldn't have found it until the spam reports arrived.

As AI makes capable software easier to build, choosing among good options becomes work. A familiar product can be worth a small premium when it gives you a reasonable place to stop looking.

AI lowers the cost of implementation. Building for yourself lowers the cost of figuring out what to build and makes software for a market of one practical.

Companies are building workflows around what today’s AI can’t do. As the technology improves, those workarounds can preserve unnecessary labor and keep better models working like the ones they replace...

Context is everything with AI. Meeting summaries are a simple example. When the task changes, go back to the transcript and ask for the slice you actually need.

I put together a deck for product and engineering leaders and barely struggled with the slides. When I dug into why, it wasn't about slides at all. I prioritize tools I can hand to an agent, and it means I get to stay in the narrative while the busywork happens around me.

A team can get to near-total AI adoption and still feel like things aren't moving. The tools changed. The jobs didn't. Here's the difference between the teams that pulled ahead and the ones who just made their old work faster.

Two things matter, context and tools. Here's how to walk yourself from a one-shot prompt to a scheduled task that runs your morning, using a kid's school inbox and tomorrow's calendar.