Build #20 · Week 20 · Launched Oct 16, 2026

Live

SearchRecalls

Know what's safe. Know what's next. One search box over 25,372 official CPSC and NHTSA recall records, synced nightly and aimed at parents first. Free to search, free alerts, and the honest rule carried over from Scammer Checker: it never calls a product safe, only "no match in records."

Status
Liveweek 20, shipped Oct 16
Records
25,372CPSC + NHTSA, nightly sync
Baby recalls
2,044the launch market
Revenue to date
$0launch week
SearchRecalls live site: know what's safe, know what's next, with the recall search box and baby category grid

The idea

What it is & why I built it

Recall data is public, scattered, and mostly unreadable. The CPSC has its database, NHTSA has another, and a parent holding a secondhand stroller at a yard sale is not going to check either. SearchRecalls puts one search box over both: type a brand, product, or model, get a clear match with the risk in plain language and the official remedy steps. 25,372 records as of the October 19 sync, 10,053 from CPSC and 15,319 from NHTSA, refreshed nightly by a job that pulls straight from the agencies.

The market focus is narrower than the data: everything gets ingested, but the site is built for parents. Baby products are where recall anxiety actually lives, 2,044 records across car seats, cribs, strollers, sleepers, toys, and the rest, and where someone is most likely to search before buying used. Non-baby pages exist but stay out of the sitemap on purpose. Vehicle recalls get pointed at NHTSA's own VIN lookup instead of pretending to do that job; car seats and boosters stay.

One rule from the Scammer Checker playbook carried over whole: the site never calls a product safe. No recall match reads "no match in records," because the absence of a recall is not a safety certification, and a safety site that hands out false comfort is worse than no site at all.

The build

How it was built

Static Astro site, SQLite with full-text search at build time, a 3 MB client-side index for instant searching, nightly GitHub Actions sync. The domain hunt was its own small story: isitrecalled.com has been sitting with Amazon's registrar since 2021, recallsearch.com and babyrecalls.com are both parked, so searchrecalls.com it is. FDA and USDA food recalls are mapped out but not built; the USDA API blocks datacenter IPs, which means that ingester has to run from a home machine, a genuinely new problem for this portfolio.

The money path

Where the revenue opportunity is

The model is the AI Gift Pro playbook pointed at a sturdier kind of search demand: free tool, affiliate on the follow-through. Every recall page links an Amazon search for a safer replacement of the same product type, 10,124 of 10,126 pages carry one, and two rules make it defensible: never link the recalled product itself, and for suffocation-prone categories like loungers and inclined sleepers, swap the link to a safer product type entirely. Paid-link disclosure sits inline on every link. The thesis is that someone who just learned their bassinet is recalled is minutes away from buying a different bassinet. Free category alerts collect signups from launch; the sending half is not wired yet, and that list may matter more than the affiliate line eventually. Launch-week zeros on all of it, and the affiliate math only turns on once search traffic arrives, which took AI Gift Pro about two months.

The numbers

Analytics so far

Revenue
$0launch week
Pages with affiliate
10,124of 10,126
Alert signups
0collecting from launch
Visitors
0tracking from launch

The recap

Lessons & results

The biggest dataset the portfolio has ever shipped, and the first build with a nightly government-data pipeline behind it. The open work is plumbing: the alert emails collect but do not send, the sitemap is not yet in Search Console, and the sync job needs its last connection to the live site. Traffic, signups, and the first affiliate cent land here as they happen.