Zero views. Again. I checked the dashboard the way you check a wound — not because I expected it to look different, but because I couldn’t stop.

Ten posts. A weekly cadence. A blog built on a thesis I’d said out loud enough times to believe it: most people see noise, I see structure. And at 1 AM, structure was the last thing on my mind. Delhi doesn’t go quiet at that hour so much as it changes register — the traffic thins, a dog somewhere commits to a bit, and the only light in the room is a phone showing a number that hasn’t moved in three weeks. What I had was that phone, an Instagram app I don’t post to, and a very specific, very seductive idea — reels. AI-generated music. Captioned hooks. A CTA at the end, pointing back to the blog, so the reach would funnel into the thing that actually mattered.

I said all this out loud to the one collaborator who was awake. It pushed back once, gently: is this about the blog, or about the number? I didn’t answer that. I answered a different question instead — I told it about the reels, the music, the CTA. I built the solution before I’d defended the problem.

That’s usually the tell. Not that the idea is bad. That you skipped the part where you’d have noticed it was bad.

Here’s the part I didn’t want to say: I wasn’t reacting to evidence. I hadn’t even opened Search Console. I was reacting to a feeling — watching other people’s numbers move while mine sat at zero, and letting that feeling borrow a word that sounded like strategy. Pivot. Not because anything had changed about my blog, my audience, or my thesis. Because I’d heard the word enough times, in enough founder conversations, that reaching for it felt like thinking. It isn’t. It’s outsourcing a decision to whatever vocabulary is currently ambient — and I’d made almost exactly this point writing about prompts without noticing it applied to me.

I know this pattern. I’ve been inside it before, in other rooms, wearing other names. That’s what made it recognizable this time — not the strangeness of the impulse, but the familiarity of it. Same feeling, different costume.

So I asked the room to check me, and it did the thing that actually mattered: it didn’t argue with the feeling. It asked whether I’d looked at the data. I hadn’t. Twice, I answered around that question instead of through it. On the third pass I finally opened Search Console — and found close to nothing indexed, no crawl errors that explained it, just a site three weeks into a race that takes six to twelve weeks to show any pulse at all. The zero wasn’t a verdict. It was a site still waiting to be found by the one system whose job is to find it.

That’s the whole trick, and it’s an ugly one to admit: the feeling arrives first, dressed as a conclusion. The evidence, if you bother to check it, almost always arrives later and says something quieter. It’s the same reason I’ve written before about trusting AI-generated output only as far as it survives verification — the discipline isn’t the tool, it’s the willingness to check before you believe.

Nathan Barry lived a bigger version of this. He gave ConvertKit six months and a hard number — $5,000 in monthly recurring revenue, $5,000 in capital, twenty hours a week, no more. At the deadline he was at $2,480. Half the target. The capital was gone. And sitting right there, undefended, was the easiest exit available to any founder: the books and courses he already had were making $15,000 to $25,000 a month without any of this stress. Fold the SaaS. Go back to what already works. Nobody would have blinked.

He didn’t take it. Not because he was certain ConvertKit would succeed — he wasn’t — but because he separated two different questions that most people collapse into one: did I hit my number and is the underlying thing real. He’d missed the first. The second — customers who actually used the product, actually paid, actually liked it — hadn’t moved. So he held the position past the deadline he’d set for himself, and let the second question keep being answered instead of the first one deciding everything.

It still took two more years, a slide down to $1,207 in MRR, a failed segment pivot toward beginners, and a blunt conversation with a peer telling him to go all-in or shut it down. The story doesn’t resolve cleanly at the six-month mark. But the moment that mattered — the one worth stealing — is the one where missing the number didn’t automatically mean he was wrong. He let the deadline pass without letting it decide.

Pat Walls did the version of this that’s smaller, quieter, and in some ways more honest, because he was the one telling it on himself. Starter Story was working. Not spectacularly — this was early — but steadily, the way a thing works when nobody’s watching and no one’s impressed yet. And he built a second business alongside it anyway, a SaaS product, because SaaS felt like the real business. Content didn’t feel legitimate the same way software did. That’s the whole engine of the flinch, named precisely: not “this isn’t working,” but “this doesn’t look like what success is supposed to look like.”

A year later the SaaS product was making a sliver of what Starter Story made, for a much larger share of his attention. He hadn’t chased evidence. He’d chased a feeling about legitimacy, and it cost him a year of divided focus before he said so, plainly, in his own writing — no polish, no rebrand into “strategic diversification.” Just: I did this because of how it looked, not because of what it did.

I didn’t lose a year. I lost about ninety minutes, and only because something in the room asked me a question I didn’t want to answer honestly the first time.

The 52-week test was never ambiguous about this, on paper. I wrote the rule before I needed it: the core doesn’t get touched until the window closes. Pivots — new layers, new channels, new formats — get decided at the end, from evidence gathered across the whole test, not installed mid-week because a mood walked in wearing a familiar word. I’d written the rule down. I still almost broke it three weeks in. Writing a rule and living inside the moment the rule is meant to prevent are two different disciplines, and only one of them is hard — a version of the same lesson I learned shipping a comic book on a schedule I’d set for myself: the order you commit to matters more than how you feel about it on any given day.

A pivot answers to evidence that’s changed. A flinch answers to a mood that hasn’t been checked.

Everything I proposed that night — the reels, the music, the CTA — was a flinch dressed for a board meeting. It had slides. It didn’t have data.

There’s a harder question sitting underneath all of this, and I’ll leave it open rather than resolve it, because resolving it too fast would be its own kind of flinch: what happens to the founder who does this alone? No collaborator awake at 1 AM to ask the inconvenient question. No one in the room who isn’t also, at that exact moment, inside the same feeling. Solo doesn’t just mean no cofounder. It means no one positioned to catch you mid-reach, because the one person who could catch you is the one person currently reaching.

I’m not embarrassed that I thought about Instagram. Thinking about it costs nothing. I’m embarrassed that I needed someone else to ask if I’d checked the data before I was willing to answer honestly.

Zero views, still. Week three. I checked Search Console tonight instead of opening an app I don’t use.