Everyone "knows" vibe-coded apps are abandoned within weeks. Someone ships a weekend project with Cursor or Lovable, tweets about it, and the domain quietly dies when the trial credits run out.
I run VibeKing, a directory of products built with AI coding tools. A couple of weeks ago I added a liveness engine: every product in the catalogue gets an HTTP check every 6 hours, cycling through all of them continuously. No sampling, no estimates — every single URL, checked repeatedly, with failure counts.
Here's what the data actually says so far.
93.1% of them are up
Of 2,587 approved products:
| Status | Count | Share |
|---|---|---|
| Live | 2,408 | 93.1% |
| At risk (intermittent failures) | 178 | 6.9% |
| Confirmed dead | 0 | 0% |
"At risk" means the checker has caught the site failing on recent checks but it hasn't sustained enough consecutive failures to be declared dead. "Dead" requires repeated failures over multiple checks — a bar nothing has cleared yet in the observation window.
Two honest caveats. First, this window is short — the engine has been running since early August, so this measures current availability, not long-term survival. The interesting numbers arrive in 6-12 months, and I'll publish those too, whichever way they cut. Second, there's a survivorship component: some products died before ever being catalogued, and those deaths aren't in the denominator. What this data can say is narrower but still useful: of the vibe-coded products visible enough to be catalogued, 93% are up right now — the wasteland narrative describes launches that never got noticed, not the ones that did.
What vibe coders build: tools for other vibe coders
Category distribution of the same 2,587:
| Category | Count | Share |
|---|---|---|
| AI Coding | 691 | 26.7% |
| Dev Tools | 400 | 15.5% |
| Marketing | 316 | 12.2% |
| SaaS | 237 | 9.2% |
| Productivity | 178 | 6.9% |
| Design | 161 | 6.2% |
42% of AI-built products are developer tooling. The community's biggest output category is tools for itself — the classic gold-rush-shovels pattern, visible in the data.
.com still wins, but the builder TLDs are a third of everything
| TLD | Share |
|---|---|
| .com | 46.7% |
| .ai | 11.4% |
| .app | 10.1% |
| .io | 6.0% |
| .dev | 3.0% |
.ai + .app + .io + .dev together are 30.5% — nearly a third of vibe-coded products launch on a "builder" TLD. And still, when makers want to be taken seriously, they buy the .com.
Method, briefly
- Catalogue: 2,587 approved products on vibeking.fun, collected from launch platforms and direct submissions
- Liveness: HTTP checks in rotating batches every 6 hours via Cloudflare Workers cron; a product is "at risk" after repeated recent failures and "dead" only after sustained consecutive failures
- Anti-bot handling: sites that block automated checks (401/403/405/429) count as alive — refusing bots is not being down
- Nothing here is estimated. Counts are from the production database, queried today.
I'll re-run this in a few months when the longitudinal picture exists. If you want your product in the tracked set: vibeking.fun/submit — free, no signup wall for browsing, and the checker will keep an eye on your uptime for you.
Top comments (0)