The pipeline was built. A product emerged.

The trend-detector pipeline has been running every morning for a week now. It's spitting out a briefing with daily BUILD / WATCH / SKIP verdicts.

At some point yesterday, I realized the pipeline wasn't the interesting part. It was the briefing. The end result. Well, not really the end result (a working app/tool people actually pay for is the actual end result). But the moment that clicked, Warranted became a project.

The pipeline pulls 400+ rising search trends, filters about 90% of the noise through is_buildable(), validates pain on Reddit, checks competition via Brave, and runs Claude to produce a confidence score with sourced reasoning. Anyone can look at Google Trends. That's easy. But reading a raw trend list isn't enough. After a week of morning briefings, when it said BUILD with a confidence score and a reason, that's the moment this became a project.

What shipped

The validation came first. Before writing a line of landing page code, I worked through five challenges to the concept:

• Sources. trendspy is the PyPI package that calls Google's internal trending_now() API. That's 400+ rising topics per call. Also, that is NOT the unrelated French dropshipping tool. The source isn't the moat. The judgment layer is.

• Cadence. Daily is the right call for this pipeline. Four consecutive days of output show meaningfully different clusters. I built the trajectory field (tracked by trend_memory.py) specifically to separate a one-day spike from a sustained trend. That field earns its place.

• Pricing. $19/month at launch. Not because the value doesn't justify $39. It does. But at zero social proof, reducing friction matters a hell of a lot more than margin. The free tier shows enough to prove the pipeline is real and signal quality is high, but withholds the actual verdict and reasoning.

• Brand. Amber #F97415 as a semantic signal color, not an accent. Reserved for BUILD labels, the primary CTA, and the key data points, never as decoration. Killing the "we" language: "Runs every morning at 6am" earns more trust than "We run this daily."

• Name. Warranted. As in: evidence-warranted confidence.

Then the landing page. Single-page Next.js, with the permission/confidence frame as the core positioning: You already know what's trending. This tells you whether it's worth building and why. Free email signup (weekly Monday digest via Resend). Paid CTA ($19/mo Stripe Payment Link). Real verdict block visible before any copy.

The pipeline is not connected. Deliberately. The landing page exists to validate demand. If people subscribe, the briefings start flowing.

Why

As soon as I built the pipeline, I had a broad idea to make it public. But I was missing the bigger picture. The valuable part was always the BUILD call, not the raw trend data. That reframe changed the positioning, the copy, the name, and the price. I didn't change a single line of pipeline code.

The hardest part of this is that I'm not testing a feature. I'm testing whether I built the right thing in the first place.

What's next

The landing page is live. The rush is on to connect everything and have a fully functioning delivery system.