Launching | data study | 7 min read
Show HN by the Numbers: What 46,138 Posts Scored in 12 Months
We pulled every Show HN post from October 2025 to September 2026. The median score was 2, one post in 60 reached 100, and timing mattered less than the title and the crowd.
- 1.1,000+ points11 posts, 0.02%
- 2.500+ points62 posts, 0.13%
- 3.200+ points332 posts, 0.72%
- 4.100+ points767 posts, 1.7%
- 5.50+ points1,373 posts, 3.0%
- 6.10+ points3,799 posts, 8.2%
- 7.2+ points30,709 posts, 66.6%
Out of 46,138 posts, how many reached each score.
What score does a typical Show HN get?
The typical Show HN post ends on 2 points. Across 46,138 posts from 1 October 2025 to 30 September 2026, half scored 2 or less, 33.4% never got past the 1 point every post starts with, and 62% had no comments at all. Only 8.2% reached 10 points and 1.7% reached 100, which is about one post in 60.
The top 1% of posts took 40.3% of all points. The top 10% took 75.9%. The bottom half shared 7.9%. If yours lands on 3 points, you had an ordinary result, not a failed launch.
How many Show HN posts go up each month?
Between 2,396 and 6,217. The count ran 2,396 in October 2025, peaked at 6,217 in February 2026, fell to 2,642 in May and settled at 3,737 in September. That is about 126 posts a day on average.
The two busiest months were the two hardest. In February and March 2026 only 1.3% of posts reached 100 points, against 2.4% in the quieter months of October and April. The Show HN rules now say it outright: "Don't post quickly-generated one-offs; anybody can do that now."
What is the best time to post a Show HN?
Weekday afternoons in the US did best by hour, and weekends did as well as or better than midweek. Posts sent between 15:00 and 18:00 UTC reached 100 points 2.2% of the time, and those sent between 06:00 and 09:00 UTC 0.8%. That hour band is real but small next to everything else on this page.
One reason to expect this: the Hacker News FAQ says rank divides points by a power of the time since submission, so early votes count most while a post is young. A smaller field on Sunday gives each new post a longer look. That is our reading of the numbers, not something the data proves.
Which Show HN titles do better?
Short and concrete ones. Titles under 30 characters reached 100 points at more than twice the rate of titles of 70 characters or more. Posts with an AI word in the title reached it half as often as those without, and they made up 34.2% of all posts.
| Title pattern | Posts | Reached 100 | Compare to |
|---|---|---|---|
| Title under 30 characters | 2,934 | 2.6% | 1.7% |
| Title of 70 or more characters | 4,743 | 1.2% | 1.7% |
| Names the browser | 1,512 | 2.6% | 1.7% |
| Names Rust | 1,035 | 2.3% | 1.7% |
| Says open source or OSS | 2,485 | 2.1% | 1.7% |
| Says I or we built, made, created or wrote | 2,865 | 2.0% | 1.7% |
| Uses an AI word (AI, LLM, agent, MCP, GPT and similar) | 15,783 | 1.0% | 2.0% |
"Compare to" is the rate for all posts, except on the AI row, where it is the rate for posts with no AI word. The title length counts the text after the "Show HN" label.
Does a Show HN need a link?
Posts with a link reached 100 points at 1.7%, nearly double the 0.9% of the 2,520 text-only posts. That fits the rules, which call Show HN "something you've made that other people can play with" and ask you to let people try it "ideally without barriers such as signups or emails".
What can this data not tell you?
"So a short title gets me more points?"
Not by itself. These are correlations. A short title may simply belong to a well-known project or a clear idea.
"And AI words hurt?"
Maybe, or AI posts are where the quickly built one-offs pile up. The data cannot separate the two. Use the word if it is what your project is, and be specific about what it does.
"Are removed posts counted?"
We took what the search API returned and did not filter further, so we cannot say how many flagged or killed posts are inside the totals. Scores are as of 31 October 2026.
"Is the weekday gap real?"
Treat the weekday gap as a hint. Sunday's 2.0% and Wednesday's 1.4% rest on roughly 100 posts each.
How should you post a Show HN?
- Link to something people can use today, with no signup in the way.
- Say what it is in under 50 characters, using words a stranger already knows.
- Skip the launch-week buzzwords. If it uses AI, name the job it does.
- Post around 15:00 UTC, on a Sunday or Monday if you have the choice.
- Stay in the thread. The rules require you to be "around to discuss" it.
- Never ask for votes. The FAQ says HN penalizes or bans submissions and accounts that do.
Then plan for the median. Two points is the likely outcome, so a launch should not rest on one post. A listing on CodeTrendy keeps working after the thread has scrolled away.
Show HN questions
How many points does a Show HN need to appear on the Show page?
HN does not publish the number. Its rules say only that a post appears on the show page once it clears "a small points threshold". Every post appears on shownew first.
Can I post the same project twice?
The FAQ allows a small number of reposts if a story has not had significant attention in the last year or so. It asks you not to delete and repost. Minor version updates generally do not qualify, though a major overhaul probably does.
Does account karma help?
The FAQ answers "No" to whether posts by users with more karma rank higher.
How we ran this study
On 31 October 2026 we queried the public HN Search API for every story tagged show_hn, in 12-hour windows so that no window hit the 1,000 result limit. That returned 46,141 records. We dropped repeats and titles that did not start with "Show HN", leaving 46,138. The newest post was 31 days old, so scores had settled. Title groups were matched by regular expression, and an AI title match is a keyword list, so it will miss some and catch a few it should not.
GET /api/v1/search_by_date?tags=show_hn&hitsPerPage=1000
&numericFilters=created_at_i>=START,created_at_i<END- Show HN guidelines, read 31 October 2026
- Hacker News FAQ, read 31 October 2026
- HN Search API documentation, the data source
Sponsored comment
Continue Reading
More articles you might enjoy
Launch
Developer Tool Pricing: Usage, Seat or Flat Fee, Chosen in Five Junctions
Five yes or no questions that route a developer tool to usage-based, per-seat or flat pricing, then the free tier shapes that fit each one, with real pricing pages as examples.
Developer Tools
Git Internals Quiz: 8 Questions on Objects, Refs and the Reflog, Answered With Real Output
Eight questions on what Git really stores: blobs, trees, commits, tags, HEAD, packed refs and the reflog. Answer each one, then open the answer to check it against real terminal output.
AI Tools
Build Log: Generating llms-full.txt, Markdown Page Copies and a Link Header From One Docs Folder
Five commits that turn a folder of Markdown docs into llms.txt, llms-full.txt and a .md copy of every page, with a check that fails the build and the v2 Link header that points agents to the Markdown.