Pen-and-ink illustration of a mesh sieve with rough lumps caught on top while fine grain falls through

I used to have two people to keep my bad ideas from mucking up our product. AI has in many ways removed both of them, but I'm actually slowly putting them back.

I get a lot of ideas and product thoughts while I'm walking my dog Bruno. For most of my career, the ones that formed a semblance of concreteness got jotted into a note on my phone. They would sit there until I could sell an engineer and a designer on building them. A lot of those ideas didn't deserve to get built, and that pitch was where the bad ones died before they cost anyone anything.

That friction was a filter, though I didn't see it that way until it was gone.

Constraints Often Increase Value

I didn't appreciate any of this at the time. Getting two other people on board did things for me that had nothing to do with whether the idea got built.

Pitching forced me to articulate the idea. To get an engineer and a designer to spend a sprint on something, I had to explain it clearly enough that it made sense to someone who wasn't already living in my head. A lot of my "great" ideas fell apart right there, the moment I had to put them into words another person could act on. If I couldn't explain why an idea mattered, it usually didn't.

The people in those roles also pushed back. They had watched me chase shiny things before, so they wanted to know why, and who it was for, and what about the part I had skipped over. Some of my favorite ideas died in that conversation, and the product was better for it.

It also took time. Pitching something could take days, and a lot of ideas die of natural causes if you give them a few days. The spark that felt brilliant while Bruno sniffed a tree he had sniffed a thousand times before, usually looked dumber by the time anyone could have started building it. None of it was deliberate, just the friction that came with needing other people.

Now It Goes Straight From My Brain to the Repo

All of it came down to the same thing: friction. That is what the last couple of years have been removing.

Now I go straight from a thought to Claude Code on my phone. Heck, I don't even have to type anymore. Plan to build to deploy, before the walk is even over. The gap between having an idea and shipping it has nearly disappeared. I've spent a career, and more than a few blog posts, on that gap, and now it closes somewhere between my front door and the corner.

I love this. I've shipped things on my own I never could have built before, in a fraction of the time. I'm not here to complain about any of it.

Claude Thinks Every Idea I Have Is Brilliant

Claude is exceptionally good at making me think every idea that sparked in my synapses is truly great. It almost never tells me an idea is bad. Mostly it just builds it for me.

So now nothing stands between the idea and the repo. Which also means nothing stands between me and fiddling with the dumb little feature I was excited about for ten minutes waiting for the sniffer to move on.

Nobody makes me defend the idea, nobody asks who it's for, and there's no wait for the shine to wear off. The idea shows up and it gets built, good or bad, because building is now easier than pausing to ask whether I should. Losing the filter didn't make my ideas any better. It just gave the bad ones a clear shot at the repo.

The Triad Did More Than Ship

This is the part I have come to appreciate more.

We talk about the designer, engineer, and product manager triad as an execution machine: design to make it good, engineering to make it real, product to decide what gets built. That's true as far as it goes, but it misses something I only saw once I was working without the triad.

The value of the product triad is far more than just execution. There is value in the friction it creates. Three people who want different things rub against each other, and that resistance catches bad ideas before they turn into bad features. That back-and-forth is the team running quality control, even when no one calls it that. On a healthy team it is some of the most valuable work going, and no one is even assigned to do it. It falls out of people caring about different parts of the same problem.

Replace that team with a tool that says yes to everything, and you lose more than the hands. You lose the part that used to tell you no, and that was worth more than I thought.

Learning to Manufacture the Friction

In the new world we live in now, I have to manufacture that friction myself. That is harder than it sounds, because every instinct I have points toward just opening Claude.

So I am trying to slow myself down on purpose. Sit with the idea. Question it. Be the skeptic the team used to be for me. In practice that looks like writing down what the idea is and why it is worth building before I let Claude touch anything, the way I would have for a pitch. If I can't make that argument to myself, I have my answer. Plenty of times I just leave the idea in the note and let the walk end, because the thing that felt urgent on the sidewalk tends to look a lot less interesting by morning. When I do build, I make myself ask what the team used to ask me. Who is this actually for? Would I have fought for it in a roadmap review, or am I building it because I can?

None of this comes naturally. I spent years resenting the friction and wishing I could just make things happen faster. Now I am building back, by hand, the friction I used to wish away.

Turns out the faster you can build, the more discipline it takes not to.