A submission with eighty points can sit on page three while one with thirty sits at the top of the front page. That is not a glitch, and understanding why is most of the work.
Hacker News launched in 2007 out of Y Combinator, built by Paul Graham, and its ranking has always been a function of points against time rather than points alone. The rough shape of the formula has been public for years: points minus one, raised to a power under one, divided by the age of the submission in hours plus a constant, raised to a power close to two. The exponents matter less than what they imply. Age punishes hard, early points count for far more than late ones, and the marginal value of each additional point falls as the total climbs.
On top of that sit penalties, which are the part nobody plans for and the reason so many good submissions vanish.
Rule One: The First Two Hours Decide Everything
Because age is the dominant term, the window in which a submission can reach the front page is short, and it closes whether or not anyone has seen your post.
A submission that gathers fifteen to twenty points in its first hour is usually on the front page. A submission that gathers the same twenty points over six hours is usually on page two and falling. The curve is unforgiving enough that a genuinely good article posted at a dead hour will often do worse than a mediocre one posted at a live one.
This is why timing advice for Hacker News is not superstition. The site’s traffic is heavily weighted toward North American working hours, and the /newest page moves fast enough that a submission posted at 3am Pacific has scrolled off before most readers arrive. Weekday mornings Pacific time, roughly 7am to 10am, put your post in front of the largest pool of people browsing /newest, which is where front page submissions are discovered.
The corollary is that you get one shot per piece, so do not spend it on a day when you are not available to engage.
Rule Two: The Title Is the Entire Pitch
Hacker News shows a title and a domain. No image, no description, no social preview. Whatever persuasion happens, happens in those words.
The title that works is the article’s own headline, stated plainly, with no adjectives and no promise. “Postgres query planner picks the wrong index at 40M rows” works. “How we achieved a 10x speedup with one simple trick” does not, and will likely be rewritten by a moderator. The audience is unusually hostile to anything that smells like a growth team touched it, and that hostility is expressed through flagging rather than argument.
Three specific title failures account for most of the damage. Numbers in a listicle shape read as content marketing. First person superlatives read as self promotion. Question titles read as bait unless the question is genuinely technical. Strip all three and what remains is usually a statement of fact, which is exactly what the site wants.
A useful test: would this title be a reasonable name for a conference talk? If yes, post it. If it reads like a blog post designed to rank, rewrite it.
Rule Three: Penalties Are Real and Mostly Invisible
The ranking formula is well known. The penalty system is not, and it is why people conclude the site is random.
Submissions can be downweighted for several reasons. A domain with a history of promotional posts carries a standing penalty that applies to everything from it. A title containing certain marketing patterns gets weighted down automatically. A flamewar detector looks at the ratio of comments to points and penalizes threads that look like argument rather than interest, which means a controversial post can be killed by its own comment section. User flags apply their own weight.
None of this is announced. A penalized submission simply underperforms its point count, which from the outside looks like bad luck.
The practical response is to stop treating the point total as your score. Watch position instead. If a submission has thirty points and is not on the front page ninety minutes in, something is weighing it down, and the cause is almost always the title or the domain rather than the content.
Rule Four: Write for the Reader Who Will Check
This audience reads the article. Not all of it, not always carefully, but enough to catch a claim that does not hold.
What that means in practice is that the piece has to survive someone with domain expertise opening it in a second tab. Numbers need a method attached. Benchmarks need their conditions stated. Any claim about a competing tool needs to be accurate, because the person who built that tool is likely reading. The comment section is a peer review process, and it is public.
The upside is that depth is rewarded in a way it is not on any other platform. A long, technical, specific piece with real detail does better here than a tight 800-word summary of the same material. Hacker News is one of the few remaining places where the correct response to a complicated subject is to write more rather than less.
What fails: anything gated, anything with an interstitial, anything where the article is a wrapper around a product pitch. Readers flag those within minutes and the flag weight does the rest.
Rule Five: Show HN Is a Different Game
If you built something people can try, Show HN is usually the better route, and it has its own rules.
The threshold is lower. Show HN submissions appear on their own list and need fewer points to surface, which makes the timing pressure less severe. The audience arrives expecting a product rather than an article, so the promotional tone that would sink a normal submission is accepted, within limits.
The limits are strict about one thing: it has to work. A Show HN with a waitlist instead of a product, or a demo that requires a sales call, gets taken apart in the comments. People will click through, try it, and report what happened. Anything that cannot be used in the next two minutes should not be a Show HN.
Write the first comment yourself, before anyone else arrives. Explain what you built, what technical decision was interesting, and what it does not do yet. The admission of limits is what buys you a fair hearing, and the technical detail is what makes the thread interesting enough to climb.
What Happens When It Flops
Most submissions get three points and disappear, including good ones, and the correct response is almost never to repost immediately.
Duplicate detection means a URL submitted again soon after will usually be folded into the original or quietly killed. The site does allow reposting after a reasonable interval when a submission got no attention at all, and the convention is to wait rather than to retry the same morning. A URL that genuinely went unseen can be submitted again weeks later with a better title at a better hour.
There is also a moderator run mechanism, often called the second chance pool, which pulls submissions that looked promising but arrived at a bad time and re-releases them onto the new page later. You cannot apply to it in any meaningful sense, but it exists, and it is one reason a post occasionally climbs a day after it was submitted.
The more useful response to a flop is diagnostic. Read your own title again cold. Nine times out of ten the problem is visible: a word that reads as marketing, a framing that centres your company rather than the subject, or a topic that was interesting to you and not to anyone else. Fix that and the next submission does better.
Rule Six: Never Touch the Votes
The thing that destroys accounts is vote manipulation, and the detection is better than people assume.
Asking colleagues, posting the link in a group chat with a request, or setting up accounts to vote together all produce the same signature: a cluster of accounts that vote in correlation. That pattern is detectable regardless of timing or IP spread, and the consequences run from the submission being killed to the domain being banned sitewide. A domain ban is close to permanent and affects every future submission from your company.
This is worth stating plainly because the advice circulates anyway. There is no safe amount of coordinated voting. The risk is not that it fails, it is that it works once and then costs you the channel.
What you may do: post the link somewhere your audience already is, with no request attached. People who find it and vote because they wanted to are indistinguishable from any other reader, because that is what they are.
The same principle applies to your own account. Submit things other than your own work, comment on other people’s threads, and generally behave like a participant. An account with history reads differently to the moderation system and to the readers checking who posted this. An account created the week of your launch, whose only submission is your own domain, is the profile of exactly the thing the site filters out.
How This Fits a Real Press Strategy
One thing worth being clear about: Hacker News is not a channel you can plan around, and treating it as one leads to bad decisions.
It is unpredictable by design. The same article can do nothing on Tuesday and reach the front page on Thursday, and no amount of preparation changes the variance much. That makes it a poor foundation for anything that needs a forecast, which is most of what a growing company needs from its communications.
What it does well is produce a credible, dated, public artifact. A thread where experienced engineers discussed your work, with your own answers in it, is evidence of a kind that is hard to manufacture. That is what makes it useful as an input to everything else, because journalists respond to external validation with specifics attached far better than to a founder’s own description.
Across our own work, which runs to more than 2,000 published articles and a journalist network above 1,200 contacts, the pattern is consistent. The pitches that convert are the ones carrying something checkable that someone else created. A Hacker News thread is one of the cheapest ways to generate that, which is a better reason to try than the traffic.
Rule Seven: The Comment Section Is the Second Submission
Reaching the front page is half the outcome. What happens in the thread determines whether the visit was worth anything.
Be there. Answer questions with specifics, concede the valid criticisms directly, and resist defending anything that does not deserve defence. The audience assigns a great deal of credit to an author who says they were wrong about something, and almost none to one who argues. A thread where the author engaged well converts readers into users at a rate that bears no relation to the raw traffic number.
Respond to the harshest technical comment first and most carefully. Everyone reading is watching how you handle it, and a precise factual answer to a sharp critique is the most persuasive thing that can appear on the page.
And expect the traffic to be a spike rather than a trend. A front page position sends a large number of visitors in a few hours, almost none of whom return. The lasting value is the thread itself, which ranks, gets cited, and becomes part of what AI assistants have read about you. Treat the day as a one-off event with a permanent artifact attached, and write the comments accordingly.