Built a daily AI tool leaderboard in a weekend. Here's what I learned about top-10 stability.
Spent the weekend building saas.pet — a free daily leaderboard that
aggregates GitHub Trending + Hacker News + Product Hunt and ranks the
top 30 AI tools.
No signup. No ads tracking. Just static HTML and a Vercel cron that
refreshes at 8:10 AM Beijing time.
The interesting finding: the top 10 is mostly the same day-to-day
(Cursor, Claude, V0, etc.), but #11-30 rotates wildly week-to-week.
GitHub is 80% signal, HN is 70% noise once you filter the
political/AI-policy stories.
Wrote 105 long-form reviews and 20 best-of guides over the weekend.
The pattern: 2000-word reviews get 3-5x more search traffic than
500-word ones.
Looking for feedback on:
- The 3-source weighting (GH/HN/PH)
- Whether to add user voting
- Best way to grow without paid ads
What would you do differently?
What stands out to me isn't the weighting.
It's that the answer to the weighting question quietly determines what the leaderboard is actually optimizing for.
That's one of those decisions that can look technical on the surface but end up shaping the product much more than expected.
Not something I'd rush to lock in.