Ten days ago I launched pickedby.ai. 275 commits later, I read my own landing page headline and had a strange feeling.
"Does ChatGPT recommend your product?"
It's accurate. It's what the tool does. But it was describing maybe 10% of what I was actually building.
The other 90% had been living in my head the whole time — fully formed, fully real — but completely invisible to anyone who landed on the site.
The sentence that was quietly killing us
Buried in the hero section was this line:
"Not for agencies. Built for creators."
I wrote it to be direct. Focused. The kind of copy that sharp indie products are supposed to have.
What it actually did: it told every D2C brand, every SaaS founder, every marketing consultant, every agency, and every future enterprise customer to leave. Not with bad intent. Just with bad framing.
When I finally saw it clearly, I deleted it immediately. That sentence had been a slow leak in the hull.
What I was actually building
Here's the vision I'd been carrying but never said out loud on the site:
Google built three products over twenty years that became the backbone of the modern internet economy. Search Console told you if Google could find you. Analytics told you how users found and used you. Ads let you pay to be found more. Together they created the most valuable advertising platform in history.
That entire stack — measure, analyze, optimize — does not exist yet for AI recommendations. ChatGPT handles 50 million shopping queries per day. AI-driven conversion rates run at 16.8%, versus Google's 1.76–2.8%. And there is no Search Console for it. No Analytics. No Ads.
"This is the 2005 SEO moment — the empty chair where Moz and Ahrefs were born."
pickedby.ai is filling that chair. In three layers.
| Phase | Google Equivalent | Our Product | Status |
|---|---|---|---|
| Phase 1 | Search Console | AI visibility score + Probe | ✅ Live |
| Phase 2 | Google Analytics | Recommendation frequency, rank tracking, benchmarks | 🔜 Q3 2026 |
| Phase 3 | Google Ads | AI-placement optimization | 🔮 2027 |
That's the real product. Not a calculator. A platform.
Why it took 10 days to say this clearly
I knew this from day one. The problem was I was solving for the wrong constraint.
Early-stage indie products are told: be specific. Don't try to be everything. Pick a lane. And that advice is right — for execution. Start with creators, earn trust, build the data moat, then expand.
But "be specific in execution" is not the same as "be small in vision." I had conflated the two. The product was starting with creators. The vision was always the entire AI economy. I just hadn't said so.
The result: we were accidentally optimizing for sounding small.
The pivot that wasn't a pivot
We didn't change the product. Not a single feature was removed or added to make this shift.
We changed the frame. The product that was "an AI visibility score calculator for digital creators" became "the Google Stack for AI recommendations." The Phase 1 entry point is still creators — that's where the need is clearest and tools are fewest. But the framing no longer closes the door on everyone else.
Internally we're calling it Big Pie («빅파이»). Small pie was the old frame: a useful niche tool. Big pie is the real frame: infrastructure for the AI economy, starting with one segment and expanding.
The hero now reads: "The AI Visibility Platform."
The subhead: "Search Console, Analytics, and (soon) Ads — for the AI era."
Same product. Completely different company.
What we built on Day 10 to lock it in
Changing copy is easy. Drifting back is easier. So we built a 4-layer enforcement system to prevent the vision from quietly shrinking again:
- Vision Brief v2.0 — a canonical document that defines the three-layer roadmap and overrides any older instructions
- Vision Gate — a 7-question checklist that must be answered before any code change. If the answer to "which roadmap layer does this serve?" is "none," the task doesn't happen
- Tagged commits — every commit carries a tag: [p1], [p2], [p3], [all], [moat-data], [moat-brand]. After a week you can run
git log --grep="\[p2\]"and see exactly how much Phase 2 work happened. Drift leaves a paper trail - Weekly audit — every Sunday, a structured self-check: what percentage of commits were vision-aligned? Did any copy narrow the audience? Is the data moat being built?
Overkill for a 10-day-old product? Maybe. But we already lost our Twitter account because we didn't build enough friction into a bad decision. Friction in the right places is a feature, not overhead.
The moat we're building right now
The most important technical task this week wasn't the landing page rewrite. It was this:
Every time we run an AI Probe — every query we send to ChatGPT, Perplexity, or Gemini checking whether a product is recommended — we now record the full result. Product ID, query template, AI source, response text, detected rank, co-recommended products, timestamp.
That dataset is the raw material for Phase 2 Analytics and Phase 3 Ads. Google's 20 years of click logs are the real moat behind Google Ads. Same principle, different era.
We started building that moat today.
What this means if you're building too
The product you're building and the vision you're building toward are not the same thing. They shouldn't be the same thing — the product is always smaller than the vision, especially at the start.
But if the gap between them is invisible to people who land on your site, you're not being focused. You're just being quiet about what you're actually doing.
Say the big thing. Start with the small thing. Both can be true at once.
We spent 10 days being quiet about the big thing. We won't make that mistake again.