Last week I asked Claude to search through an archive of my newsletter and blog posts, looking for topics worth revisiting. It spun up a background process to do the search. That process burned through my entire token budget before it finished.
This was annoying, but it made me realise something: I didn’t actually need that search. I needed to draft a post. And I couldn’t do that either - not because the tool was gone, but because I’d got distracted with “what if” mode when I should have been focused on writing the next post.
The same pattern shows up in my side projects
There are two versions of this I’ve seen when building side projects.
The product version: building ideas too quickly. AI makes building cheap - enough so that you can build before deciding if this is really what you want. Most of what gets thrown away wasn’t wrong. It just needed a couple more iterations to get it right.
The technical version: losing track of code changes. This is where changes land faster than I can understand them. Code goes live before I’ve fully digested it. This isn’t a major issue when it happens occasionally. But over time, this problem compounds.
AI makes it easy to run into these issues - because it removes the friction that you get when going slower. Normally, if you’ve got 10 new ideas, you have to pick one - the opportunity cost forces a decision. With AI, you can build them all. That’s both a good thing and a bad thing.
Like this post so far? Product Notes is a free weekly newsletter for product people. Subscribe to get it in your inbox.
This isn’t an argument against AI
AI has allowed me to make progress far quicker than I could have done without it. That’s a real benefit.
The problem isn’t using AI. It’s letting a good outcome be the reason you stop exercising the thinking that produced it. Every time it gets something right, it’s tempting to hand over a little more of the front end - the deciding, not just the doing. Do that enough times and you get out of practice, and you don’t notice until the day the tool isn’t there.
The question I’m asking myself now
Before I hand something over now, I ask one question: could I still do this if the tool wasn’t here today? Not whether I’d want to - just whether I could.
A no isn’t a reason to stop using AI for it. It’s a flag that you’re out of practice - and that’s exactly the moment a crisis will test you on it.
I’ve kept asking myself that question since.
Product Notes is free and lands every week. Subscribe to get it in your inbox.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.