Two weeks ago I published a dataset of every Show HN post for the last 24 months — 70,894 launches — and the finding people kept quoting was the flat one: the median Show HN post earns about 1 point and zero comments, and that median didn't move in two years. A reader asked the obvious follow-up: fine, the median is flat — but what about the winners? Did the 10-point group thin out too?
I ran it. They did. And it's worse than the ticket-metaphor suggests.
The numbers (2025-H1 vs 2026-H1, both full windows, same dataset)
| 2025 H1 | 2026 H1 | change | |
|---|---|---|---|
| Submissions | 12,449 | 25,968 | +109% |
| Share reaching 10+ points | 13.0% | 7.8% | −40% |
| Share reaching 100+ points | 3.49% | 1.67% | −52% |
| Median points within the 10-point club | 39 | 30 | −23% |
Same front page, roughly the same number of slots. Twice the tickets. Half as many launches clear 100 points — the "real outcome" bar — and the launches that do clear the 10-point bar are clearing it by less. In 2025 a typical winner landed at 39 points. In 2026 the typical winner lands at 30.
The 2024-H2 cell in my data is a 10-week partial (the pull starts 2024-10-21, n=3,660), so I use the two full windows for every comparison — but for what it's worth, the partial cell points the same way and higher: 14.5% reached 10 points, 4.26% reached 100.
What this actually means
The first piece's finding was "your odds didn't move" — the median post is still invisible. This is the second shoe: the middle of the distribution is hollowing out, not just the floor. It's not only that most posts die quietly; it's that the outcomes between "died" and "front page" are vanishing. In one year, the share of launches reaching the 10–100 point band (the zone that used to feel like traction) fell from 9.5% to 6.1%, while the 100+ zone halved. The distribution is splitting into two clusters: invisible and front-page, with nothing much between.
I don't have a clean causal story for why submissions doubled — more AI-assisted building, more launch culture, more directories telling people to "launch everywhere" — and I want to be careful not to dress up a correlation as one. What I can say is mechanical: the front page is zero-sum, submissions doubled, and every slot someone wins is now won against twice the field. The winners are winning by less because the bar didn't move but the crowd did.
Caveats, in the interest of honesty
- Points are one snapshot taken at crawl time, not full histories. Late-blooming posts are undercounted; that data stopped being collectable (it's the subject of my other piece). The direction across the settled windows is unambiguous, the exact cells will drift slightly.
- 2026-H2 is partial and still accruing votes, so it's depressed by construction. I exclude it from all comparisons above.
- HN moderation (flags, kills) is invisible to a crawler. The 100+ club numbers are floors, not ceilings, on what the community "wanted" — some of what died was killed, not downvoted.
- The 2024-H2 cell is 10 weeks, not 6 months. Treat it as directional only.
The honest takeaway
If you're launching on Show HN in 2026, the median outcome is unchanged — still ~1 point, still no comments. What changed is everything above the median: twice the competition for the same slots, half as many 100-point outcomes per launch, and a middle band that used to read as "at least some traction" fading into the floor. Landing on the front page still happens — 1.67% of launches did it this year — but it is now twice as rare, relative to the field, as it was in 2025.
The dataset and analysis scripts are public as before; the 24-month pull, the by-half computation and the winner-median check all run from the same CSV. Numbers in the table reproduce from the raw data with a 10-line script.
Top comments (0)