The Shitty First Step
In her classic book on writing, Bird by Bird, Anne Lamott gave writers permission to write terrible first drafts. The shitty first draft, she argued, is how all good writing actually gets made. It is “the child’s draft,” she writes, “where you let it all pour out and then let it romp all over the place, knowing that no one is going to see it.” The goal of the first draft isn’t to be good. It’s to exist.
I’ve been thinking lately about how this same principle applies far beyond writing. There’s a parallel concept I’ve started calling the shitty first step, and it’s become one of the most useful mental models I’ve picked up over the years.
So much of what stops us from doing things isn’t the difficulty of the task itself, but the inertia of not knowing where to begin. I’ll have a home project or a coding problem or some life admin task sitting on my list for weeks, sometimes months, not because it’s particularly hard but because I genuinely don’t know what the first move is. The overwhelming feeling of I don’t even know where to start is often enough to stop me entirely.
A few weeks ago, a leg broke off one of our dining room chairs. In a previous era of my life, this chair would have sat in the corner of our dining room for months. I’d look at it occasionally, feel a vague sense of guilt, and then continue not fixing it because I didn’t know what kind of fastener or clamp I needed. Instead, I snapped a picture of the chair leg, the bolt, and the bracket on another chair, sent it to ChatGPT, and talked through what had happened.
Within a minute or two, I had enough information to know what I was looking for. Was the answer perfect? Probably not. But it was enough to get me to the hardware store with some idea of what to ask for, and that made all the difference.
This is the shitty first step in action. You’re not asking AI to solve your problem perfectly. You’re asking it to give you something, even if that something is partially wrong, so that you have a starting point to work from.
By the way, you don’t need AI for this, but it sure can help when you don’t have a human around. I’m not someone with deep knowledge of woodworking, plumbing, or electrical work. A lot of the time, I don’t know what to Google, and I don’t have a friend or family member I could easily ask who would just know. That’s where AI can come in handy.
Why wrong answers are still useful
Setting aside the means of solving the problem, the point here is that sometimes you don’t need a perfect response. You just need a response.
When I describe a problem poorly to an LLM and get back an answer that’s not quite right, that wrong answer becomes a catalyst. It gives me something to react to. I can look at it and think, no, that’s not quite it, but actually the real issue is more like this… and suddenly I’m unstuck. I’ve gone from staring at a blank wall to having a conversation, even if that conversation started with a misunderstanding.
This works because so many problems have a vocabulary barrier. I’m currently trying to troubleshoot a leak into my garage and some broken tile in my bathroom. I don’t know the right terminology for what I’m looking at. I don’t know what questions to ask. But I can take a picture and describe what I’m seeing in plain language, and even if the first response isn’t exactly right, it gives me enough context to have a real conversation with someone who actually knows what they’re doing.
That’s maybe the most underrated part of this whole approach: the shitty first step can not only get you unstuck, it can also give you enough vocabulary and context to talk intelligently to a friend, a family member, or someone at the hardware store (or whatever the right store is for you). If often feels like we need to be an expert when troubleshooting and asking for help, but really we just need to be a little less than completely lost.
Beyond home repair
I’ve found this pattern everywhere once I started looking for it. Health and ergonomics is a good example. I’ve dealt with RSI and wrist issues over the years, and I’ve used AI to troubleshoot my desk setup. I’ll snap a picture of how I’m sitting, describe where the pain is, and ask what might be causing it. This isn’t a replacement for seeing a doctor or physical therapist, but it’s a way to get unstuck when I’m not even sure what kind of professional I should be talking to or what questions I should be asking them.
Even rabbit holes of curiosity benefit from this. I was scrolling TikTok recently and fell in love with a cover of a Japanese song I had never heard before. Screenshot, send to ChatGPT, and suddenly I’m learning about a Japanese emo band I’d never heard of called Singer’s High, reading translated lyrics for their song 朝を待つ, and going deep into a whole new corner of music. That discovery wouldn’t have happened if I’d had to figure out how to search for “that Japanese song from the TikTok with the girl playing guitar” on my own. (It’s a really, really good song, by the way.)
The shitty first draft for action
Anne Lamott’s insight was that perfectionism kills writing before it starts. You have to give yourself permission to be bad in order to eventually be good. The same thing is true for action more broadly. The fear of not knowing the right first move keeps us frozen, and the antidote is accepting that the first move doesn’t have to be right. It just has to be a move. I described this as just pick something in my article How to Finish What You Start.
AI has made the shitty first step dramatically easier to take. You can describe a problem badly and still get something useful back. You can ask a dumb question without feeling judged. You can get just enough information to break the inertia and start moving.
The chair is fixed now. The garage leak is still in progress, but I’m actually making progress on it instead of avoiding it. And I’ve got a new favorite Japanese band. None of these required AI to give me perfect answers. They just required something to push against, a foothold to start climbing from.
Keep an eye out for the problems sitting on your list not because they’re hard but because you don’t know where to start. That’s where the shitty first step can help the most.