I built something in a month that would have taken a small team half a year. oven ai
Then... it wasn't actually that bad. It wasn't a total ghost town. I kept it running for two years and eventually scraped together 2,000 users. The product worked. The product was, I thought, good. But it was never quite enough to sustain a full-time business. It just sat there, plateaued, paying for its servers but not my rent. That was when I stopped being able to pretend the hard part was still the building.
The part that got easy
For most of the last decade, the honest answer to "what's your moat?" could be execution. Not because execution was noble, but because it was genuinely scarce. Turning an idea into working software required people who were hard to hire, months that were expensive to fund, and a tolerance for detail that most teams lacked. If you could actually ship, that was, for a while, enough to be a business.
That scarcity is gone. An agent writes the migration, the API route, the tests, and the landing page. A founder with taste and a terminal now covers what used to need three engineers and a designer. The gap between "I have an idea" and "it's live on a domain" collapsed from months to an afternoon.
This is wonderful. It's also the whole problem.
Because a moat is not "this was hard for me." A moat is "this is hard for the next person." When the cost of building your product drops by 10x, it drops by 10x for everyone who wants to copy it — including someone who has never heard of you and will ship your feature set next Tuesday because they saw a screenshot.
The thing you shipped is no longer defensible. It's the table stakes.
What I actually did instead
I went back to selling. Not "growth." Not "marketing strategy." Selling — the unglamorous, direct version.
I wrote to people one at a time. I asked what they were doing instead of my product, which is a better question than whether they liked it. I sat in the discomfort of watching someone use the thing while saying nothing. I wrote posts explaining what I'd learned, published under my own domain, and watched most of them do nothing for weeks.
Here's what surprised me: selling made the product better than another week of building would have. Every conversation cut something. The features I was most proud of turned out to be the ones nobody could describe back to me. The thing people actually wanted was smaller and more boring than what I'd built, and I only found that out by trying to sell it.
Building in isolation feels like progress because the feedback loop is instant — the test goes green, the page renders. Selling feels like failure because the loop is slow and the answer is usually no. That asymmetry is exactly why founders keep choosing the terminal. It's the more comfortable room.
Selling is not a phase you graduate from
The mental model I had was sequential: build the thing, then go sell the thing. Finish, then distribute.
That model is wrong, and it's expensive. Distribution isn't the step after building; it's a parallel track that starts on day one and never ends. If you wait until the product is "ready," you arrive at the market with a finished product and zero audience, which is the worst possible trade — you spent the scarce resource (time) acquiring the abundant asset (software) and none of it on the scarce one (attention).
The founders doing well right now aren't the ones who shipped best. They're the ones people already knew about when they shipped.
So what is a moat now?
If it isn't the code, it's some combination of these — all of which take time, which is the one input agents haven't compressed:
- Distribution. An audience that already reads you, a domain that already ranks, a list that already opens. This one compounds and can't be cloned in a weekend.
- Being the answer. Not just ranking in search, but being the source that gets cited when someone asks an AI the question your product solves. That's earned by publishing things worth citing, over and over.
- Trust and specificity. Knowing one narrow customer so well that your product feels hand-made for them. Generic gets copied instantly; specific doesn't.
- Accumulated context. Data, workflows, and history that make leaving expensive because the thing has learned the customer's shape.
Notice that three of those four are downstream of publishing consistently. That's not a coincidence, and it's why I ended up building an SEO product: the moat I most needed was the one I was worst at maintaining by hand.
- Consistency beats quality; therefore, why not implement an agent!!!
The ratio I got wrong
I used to spend something like nine-tenths of my week building and the rest on "marketing," which mostly meant tweeting into a void and calling it distribution.
Flipping that ratio is the single highest-leverage change I've made. Not because building stopped mattering — I still ship most days — but because an hour spent making the product better is worth very little if the number of people who could notice the improvement is zero. Multiply anything by zero and you get the same answer.
The honest version: if you can't name the last five people who found you and how they found you, your problem is not the product.
What this looks like in practice
Three things, in order of how uncomfortable they are:
- Talk to people who haven't heard of you. Not your existing users, who are already convinced. The cold ones. That's where the objections live.
- Publish on a domain you own. Every post is an asset that keeps working; every platform post is rent. More on that in SEO is the only marketing that compounds.
- Make the loop survivable. The reason founders stop publishing isn't disbelief — it's that week eleven arrives and nothing has happened yet, and doing it by hand is exhausting. Pick a cadence you can hold when it's boring. See How to get found in the era of vibe coding.
The uncomfortable conclusion
Shipping used to be the moat because shipping was hard. Now it's the entry fee. The moat moved to the thing that was always harder and less fun: getting known, staying known, and being the answer when someone goes looking.
I didn't want that to be true. I'm a builder; the terminal is my comfortable room. But the market doesn't grade on which part you enjoy. So I went back to selling. And then — because I am still, incurably, a builder — I built something to make the publishing half run without me.
I implemented Grove AI to auto-post my blog content, turning the relentless grind of consistency into an automated system. And it actually worked. Traffic climbed, views compounded, and my SEO finally started doing the heavy lifting of distribution for me.
The terminal will always be the more comfortable room. It’s where the feedback loop is instant, and the tests go green. But the moat isn't in the code anymore. It’s in the messy, unglamorous work of getting people to care. Software is abundant now; attention is scarce. Stop polishing features in a vacuum, close the editor, and go spend your time on the scarce thing.
grove researches, writes, quality-gates and publishes to your own domain on autopilot, so the distribution half of the job keeps running on the weeks you're heads-down building. Plant your domain and see the first post in the same session.
