Vol. I  ·  № 1 Apr 29 · 2026
An almanac of Product Hunt featured launches
Six months · 3,311 launches · 27 weeks

Half of launches say "AI." One category is actually growing.

Six months of Product Hunt's featured-launch tier, regressed with 95% confidence intervals. Most of what you've heard about PH trends doesn't survive the test.

3,311 posts 100 enriched 20 deconstructed 14 categories regressed
1/14
Categories actually growing
Only developer-tools has a slope whose CI₉₅ excludes zero. The pickaxe is the trend.
5/20
Top-20 launches with zero reviews
High votes, no discussion. Pattern, not proof — but the leaderboard is partly performative.
2×
Sunday vs Tuesday upvotes
Median 195 on Sun, 108 on Tue. Day-of-week is the most underrated launch lever.
§ 01 — What's actually winning

Pickaxes. Only pickaxes.

After regressing 27 weeks of weekly topic share, developer-tools is the single category with a positive trend whose 95% CI excludes zero. AI feels like a wave; statistically it's a plateau.

All 14 tracked categories · weekly share Nov→Apr
Highlighted = stat. significant slope (CI₉₅ excludes 0)
AI vs developer-tools
Weekly share · OLS fit · the only meaningful divergence
AI's slope is +0.26 pp/wk with CI₉₅ [−0.23, +0.74] — includes zero. Devtools at +0.36 pp/wk, CI₉₅ [+0.08, +0.64] is the only confident upward trend.
Tagline themes — median votes
Vertical & Claude-specific outperform generic AI
Vertical-vibe (150) and Claude-specific (138) clear the field. Generic agent theme has 526 launches at median 130 — saturated. Eval/quality theme: only 26 launches — whitespace.
✦   ✦   ✦
§ 02 — When to launch

Day-of-week is the most underrated lever.

Same product, same week. Sunday gets ~2× the Tuesday median. Yet most launches default to Tuesday.

Median upvotes by featured day
All 3,311 featured posts · across 27 weeks
Sunday: 195v median. Tuesday: 108v. Tue-Fri are the high-volume, low-yield days — most of the field crowds those slots.
Monthly upvote distribution
p10 · p25 · p50 · p75 · p90 · p99 across six months
The p99 ceiling dropped (848 → 587) but the median is mostly stable once you account for vote maturity. The bar isn't rising; the field is just thicker.
✦   ✦   ✦
§ 03 — Top 20, deconstructed

Three molds. Plus four flags.

Vertical AI with a workflow. Pickaxes for AI builders. Foundation-model & big-label launches. And a small group whose engagement metrics don't match.

The top 20, mapped to category

✦   ✦   ✦
§ 04 — Engagement integrity

When upvotes outrun reviews.

Each dot is a top-100 product. Healthy launches scale votes and reviews together. The shaded zone — high votes, zero reviews — is where the integrity flags sit. Pattern, not proof.

Votes vs reviews
Top-100 products · log scale on reviews axis
Inside the flag zone: 25 launches. Compare with Cowork (753r/4.97) and Claude Opus 4.6 (753r/4.97) on the right edge.
Flagged shape · 12 examples
High votes · zero reviews · low comment ratio
✦   ✦   ✦
§ 05 — Graveyard

What top-50 has that bottom-50 doesn't.

Random sample of 50 bottom-decile launches vs the top 50. Most folk wisdom is myth-busted; the only meaningful gap is whether the tagline mentions AI.

Tagline characteristics
% of taglines · top vs bottom 50
The only meaningful gap is AI mention (24 pp). Specific verbs are rare in both groups (the "use specific verbs" advice is sample-thin in practice).
Bottom-decile sample
Featured-but-ignored · 15 of 50
✦   ✦   ✦
§ 06 — Playbooks

What the data says to do.

For · Solo founders

Ship Sunday.

  • Sunday: median 195v. Tuesday: 108v. Day choice nearly doubles your floor.
  • Realistic target: 200v. That's the p50 of featured. 1000+ is top 1%.
  • Seed reviews first. 0 reviews + high votes reads inauthentic.
  • "AI" in tagline = table stakes. 56% of top-50 use it. Don't hide it.
For · Pickaxe sellers

Build for builders.

  • Devtools is the only growing category. +0.36 pp/wk, CI clears zero.
  • Claude-tagged products: median 138. Cowork, KiloClaw, Kilo Code Reviewer.
  • Infra/MCP/SDK floor: 123. Less glamorous; high floor.
  • Eval/observability: n=26. Demand >> supply on PH. Whitespace.
For · Indie hackers

Middle is alive.

  • No-code app builder: median 133. n=231. Saturated; pick a vertical.
  • Voice/multimodal: median 131. n=207. Speech, video, image still working.
  • Memory/context: median 117. n=81. Niche but consistent.
  • Avoid marketing & design tools. Negative point estimates, AI-eaten.
For · YC applicants

Don't launch on YC week.

  • YC week (Mar 9): 60+ tagged launches. You'll be one fish in a school.
  • Launch the week after. Quieter. Same algorithm.
  • Anchor on p75 (200-260), not 1000. Only 5 hits 1000+ in 6 months.
  • YC partners click reviews. 800v + 0 reviews looks worse than 400v + 80 reviews.
✦   ✦   ✦
§ 07 — What this is not

Caveats, with confidence stamps.

Vote maturity — the methodological footnote
Recent posts have had less time to accumulate votes
Posts >90d old: median 143v. <7d old: 88v. The "April collapse" headline (−36%) shrinks to roughly −12% once you correct for age. Important caveat, boring story.
Survivorship Acknowledged
Featured posts only (~25/day). Unfeatured submissions invisible.
Vote maturity Quantified
>90d: median 143. <7d: 88. Explains most of the "April collapse."
Self-tagging Cross-checked
Makers over-claim "AI". Tagline-text NLP cross-verified.
Event spikes Filtered
YC W26 (Mar 9), Pitch Berlin/London (Apr 20-27) called out separately.
International Section killed
99.8% English. No real non-English cohort on featured tier.
Integrity flags Pattern, not proof
Some products legitimately didn't seed reviews. We flag the shape only.
Regression power Underpowered
27 weeks. Subtle trends invisible to CI₉₅. More months = sharper.
Methodology Sourced
Adapted from Google's "Practical advice for analysis of large datasets" — confidence on every estimate, watch for mix shifts, three-stage validation.
The Hunt  ·  Vol. I  ·  № 1  ·  ~/ph-trends/report.html
Generated · local SQLite source