Questions worth following a little further.

What if you could query Surfline in natural language, customized for your own breaks and surfboards? I decided to give it a try based on my own quiver of funky twin fins and alternative shapes after TypeSafe’s Jev model made it easy to make hundreds of classifications in seconds.

A matrix I adapted from James Altucher’s idea math and ran against a timer. Put a few moving parts in the pile and five operations combine them into more ideas than you can use. Bad ideas are the point: they are what a good one arrives hiding behind.

Just because you can doesn’t mean you should. I’m testing ways to find the gems hidden in the slop.

AI is strongest in bounded work. This note explores the workflows, interfaces, and trust that could bring it to everyday people, starting with a system that knows when to reach out for help.

In 2021, a set of commissioned sketches gave a half-formed customer signal a concrete shape. Five years later, one of them was waiting when the product finally needed it.

We brought our best dispatching technology to a new market. The operators had that part handled. What they couldn’t do was fill a thin week or move their prices.

When a system is too big to fit in a prototype, the challenge becomes creating a memorable story.