Hacker News for Founders: Participation That Works

The version of Hacker News marketing that works is not posting your launch and hoping: it is finding threads where your specific experience answers a question someone already asked, then writing a comment that would be useful even if you had no product. The mechanical part of that is easy, because the Algolia HN Search API is free, needs no key, and lets you query every story and comment on the site by keyword, date, points, and author. The hard part is tone and timing, and both are learnable.

This guide covers the three things that actually determine outcomes: how to find the right threads with the API, how to write in a register the community accepts, and when to post so the ranking formula does not bury you. It assumes you are a founder or an early marketer at a technical company, that you have limited hours, and that you would rather do ten good comments a month than a hundred bad ones.

What Hacker News marketing actually is, and what it is not

Hacker News has no advertising product, no sponsored placement, and no influencer tier. There are exactly three ways your company can appear: someone submits a link to it, you submit a Show HN, or you comment somewhere as a person who happens to work there. Everything else people call Hacker News marketing is a variation on those three, and two of them depend on other people's goodwill.

The site's guidelines are short and worth reading in full, but the operative lines for a founder are these: do not ask for upvotes, do not use multiple accounts, do not post promotional filler, and be a person rather than a brand voice. Voting rings get accounts banned, and the moderators are unusually good at detecting them because vote patterns are trivially clusterable. Assume anything that looks like coordinated promotion will be caught and that the penalty is silent: your submissions simply stop reaching the front page and you never get told.

What does work is narrow and repeatable. You watch for threads on topics where you have first-hand operational detail. You show up with numbers, failure modes, and specifics that a generalist could not produce. You mention your product only when it is the direct answer to the question asked, and you disclose the affiliation in the same sentence. Over months this produces a small, durable stream of technical readers who arrive already predisposed to trust you, which is worth more than a spike of curiosity traffic that bounces in nine seconds.

The failure mode is treating HN as a distribution channel rather than a room. A distribution channel takes your message and delivers it. A room notices when you only speak about yourself.

How the Algolia HN Search API finds threads worth answering

Algolia runs the search backend for Hacker News and exposes it publicly. The base endpoint is https://hn.algolia.com/api/v1/search for relevance-ranked results and https://hn.algolia.com/api/v1/search_by_date for reverse-chronological results. No authentication, no key, no signup. There is a published rate limit, so cache aggressively and poll on a schedule rather than in a loop, but for a founder watching a dozen keywords it is effectively unlimited.

Four parameters do almost all the work:

• query: the free-text search string.
• tags: filters by object type. Useful values are story, comment, show_hn, ask_hn, front_page, author_<username>, and story_<id>. Combine with commas for AND, with parentheses for OR, as in tags=story,(show_hn,ask_hn).
• numericFilters: comparisons on created_at_i (Unix timestamp), points, and num_comments. For example numericFilters=created_at_i>1723000000,num_comments>10.
• hitsPerPage and page: pagination.

The response gives you objectID (which is also the HN item id), title, url, author, points, num_comments, created_at_i, and for comments the comment_text and story_id. Reconstructing a link is just https://news.ycombinator.com/item?id=<objectID>.

Two caveats. The Algolia index is a search index, so there is a short lag between something being posted and being findable. If you need the live state of a thread, the official Firebase API at https://hacker-news.firebaseio.com/v0/ returns item and user objects in real time, including the /v0/newstories and /v0/topstories id lists. The practical pattern is Algolia for discovery and Firebase for freshness. The second caveat is that comment search returns HTML-ish text, so strip tags before you feed it to anything.

Query recipes that turn the API into a daily watchlist

The point of the API is not one-off searching. It is a standing set of queries you run every morning that produce a short list of places where a comment from you would be welcome. Here are the patterns that earn their place:

Run the first three every weekday morning. Run the brand and competitor queries daily and treat every hit as something to read, not necessarily something to reply to. Run the Show HN precedent query once, before you launch, and read fifty threads.

The competitor query deserves emphasis. Hacker News comments are one of the few places where technical buyers compare tools in public with no incentive to be nice. Reading every comment that names your category is the cheapest product research available, and it pairs naturally with tracking how the same category gets described by AI assistants, which we cover in brand mentions monitoring in the AI era and AI visibility tracking: how to measure it.

How to read a thread before you type anything

Most bad HN comments are bad because the author skimmed. Before writing, spend five minutes establishing three things.

First, what is the actual claim being argued? HN threads drift. A story about a database release becomes an argument about licensing, then about hiring, then about a tangent someone raised in the fourth comment. Find the sub-thread your knowledge fits, not the headline.

Second, has your point already been made? If it has, and it is sitting at the top with a hundred points, adding a lower-quality version of it earns nothing. Reply to it with the additional detail instead. A reply that extends the best comment is read by everyone who read that comment.

Third, what is the sophistication level in this particular thread? HN is not one audience. A thread about Kubernetes operators has different assumed knowledge than a thread about pricing psychology. Matching the level matters more than being impressive.

Then check the clock. A comment posted eight hours after a story hit the front page will be read by almost no one, regardless of quality. The value of a comment decays roughly with the story's own attention curve. If the thread is cold, save the material. It will come around again, because HN topics are cyclical and your standing queries will catch the next instance.

Tone: what gets upvoted and what gets flagged

The community guidelines ask for curious, substantive conversation and explicitly discourage snark, name-calling, shallow dismissals, and rhetorical questions used as attacks. That sounds like generic politeness advice. It is not. It is a specific register, and you can learn it by imitation.

Concrete rules that hold up:

Lead with the specific detail, not the framing. "We ran into this: our p99 went from 40ms to 400ms once the table passed a few hundred million rows, and the fix was X" outperforms "Great post! We've seen similar issues at scale."

Name the tradeoff you accepted. HN readers are suspicious of unqualified recommendations. Saying "this approach costs us more in storage and we decided that was fine because Y" reads as honest in a way that no amount of enthusiasm does.

Disclose in-line. If you work on the thing you are describing, say so in the sentence where you first describe it, not in a footnote. "Disclosure, I work on this" is enough. Undisclosed promotion is the single fastest way to get flagged, and people will check your comment history.

Do not argue with downvotes. Complaining about votes is against the guidelines and it never helps. If a comment goes negative, leave it.

Do not paste marketing copy. The register of a landing page is instantly recognizable and instantly penalized. If you would put it on a pricing page, rewrite it before it goes in a comment.

Answer the question that was asked. If someone asks how you handle rate limiting and your answer pivots to your product's roadmap, you have lost the room.

A useful calibration exercise: pull a hundred comments from a user whose comments consistently sit at the top of threads in your area, using tags=comment,author_<username>, and read them in one sitting. The pattern will be obvious within twenty comments, and it is almost always the same pattern: short, specific, first-hand, hedged where uncertain, and free of adjectives.

Timing, and what the ranking formula implies

The widely published approximation of the original HN ranking code is:

score = (points - 1) / (age_in_hours + 2) ^ 1.8

with moderator penalties and other adjustments layered on top. You should not treat it as exact, because the real implementation has changed and includes penalties for certain domains, for titles that look like clickbait, and for threads that turn into flamewars. But the shape of it is stable and it tells you two useful things.

First, early points matter disproportionately, because the denominator is small. A story that gets its first several upvotes within the first hour has a materially better chance than one that accumulates the same votes over six hours. This is why the guidelines forbid asking for votes: the mechanism is exploitable and the moderators know it.

Second, decay is steep. The exponent of 1.8 means a story roughly halves in ranking value every few hours even with no new votes. Front page tenure is short. Plan for a spike that resolves within a day rather than a slow build.

The practical timing question is when to submit. The honest answer is that the folk wisdom about optimal posting hours is worth less than your own measurement, and the API lets you measure it. Pull every story in your category from the last year with search?query=<topic>&tags=story&numericFilters=points>30, bucket created_at_i by hour of day and day of week in your target audience's timezone, and look at where the successful posts actually cluster. That takes an afternoon and replaces guesswork with data specific to your niche, which is not the same as the site-wide pattern.

Two structural notes that matter more than the hour. The /newest page moves quickly, so a submission that gets no attention in its first stretch there is effectively gone. And if a good post genuinely got missed, moderators run a second-chance pool that re-surfaces submissions, and you can email them about it. That is a legitimate path, unlike asking friends to vote.

Show HN: the one place a product post is welcome

Show HN is the sanctioned exception. The rules for it are published separately from the main guidelines and they are specific: it must be something people can actually try, not a signup page, not a waitlist, not a video. It should be your own work. You should be present in the thread to answer questions.

What separates a Show HN that lands from one that does not:

The text field is where founders under-invest. A short paragraph explaining the technical decision that was genuinely difficult, plus an honest statement of current limitations, does more for credibility than any amount of positioning. HN readers reward people who say "this does not handle X yet and here is why that is hard."

Prepare for the traffic. A front page Show HN sends a burst of technically literate visitors who will open dev tools, read your HTTP headers, and notice if your page takes four seconds to render. Run a performance and on-page check before you submit, not after. Skopx Site Health pulls Lighthouse scores from Google PageSpeed Insights, real-user Core Web Vitals from CrUX, and Search Console performance, and produces a 0 to 100 on-page audit with a fix list, which is a reasonable pre-launch gate. If you want the underlying concepts first, see Core Web Vitals monitoring and what to look for in an SEO audit tool.

Also decide in advance what you will do when the thread is over. Front page attention is a single day. The signups it produces are worth something only if you have somewhere to put them.

Building the monitoring loop without living in the API

The workflow described above is a daily ritual: run six queries, read the hits, decide which two or three deserve a reply. That is fifteen minutes if it is automated and forty-five if it is not, and the version that is not automated stops happening in week three.

The mechanical version is a small scheduled job that hits search_by_date for each of your watch terms, filters out anything you have already seen by objectID, and posts the new hits to a channel you actually read. Skopx handles this shape of task through chat-built workflow automations: you describe the queries and the schedule in plain language, and the run produces a digest on the cadence you set. The AI Visibility feature does a related job automatically, surfacing live Reddit and Hacker News threads where your category is being discussed, alongside share of voice in AI answers and citation gaps where competitors get named instead of you.

Whatever you build it with, hold two lines. The job finds threads. A human writes every comment. Automated comments on Hacker News are against the rules, trivially detectable by tone, and will get the account banned. The automation exists to solve the discovery problem, which is genuinely tedious, not the writing problem, which is the entire value you are contributing.

If you are stitching HN monitoring into a broader distribution routine that also covers other communities and social networks, the mechanics of scheduling and adapting content per platform are covered in automated social media posting and the cross-posting tool guide. Skopx Social Autopilot handles the publishing side across LinkedIn, Facebook Pages, Reddit, Instagram, X, Threads, Bluesky, Mastodon, Telegram, Discord, an email newsletter through your own Resend account, and the Skopx community feed, with content adapted to each network's character limit. Hacker News is deliberately not on that list, and should not be: it is a place you show up in person.

Measuring whether any of it worked

HN traffic is easy to see and easy to misread. Referrer data will show news.ycombinator.com, but a large share of clicks arrive with no referrer because of privacy settings and apps, so treat the visible number as a floor.

More useful signals, in rough order of value:

1. Signups or trials attributable to the day, compared against your baseline. Blunt but honest.
2. Direct traffic lift in the following weeks. People who read a good comment often come back later by typing the domain.
3. Search demand for your brand name. Search Console impressions on branded queries after a front page appearance tell you whether the name stuck.
4. Downstream citations. HN threads get indexed, quoted, and increasingly get surfaced inside AI-generated answers, so a strong thread can keep producing months later. That mechanism is the subject of the generative engine optimization guide.
5. Comment karma trajectory on your account. A slow, steady climb means your register is right. Flat or negative means it is not.

What not to measure: single-thread upvote counts as a proxy for business outcome. A submission that reaches a hundred points and sends bouncing traffic is worth less than a single comment in a niche thread read by four hundred people who all have the exact problem you solve.

Set the expectation accordingly. Participation here is a compounding, low-volume channel. Six months of showing up produces a reputation that makes the seventh month's launch land. Six days of it produces nothing, which is why most attempts at Hacker News marketing are abandoned before they could have worked.

Frequently Asked Questions

Do I need an API key to use the Hacker News search API?

No. The Algolia HN Search API at https://hn.algolia.com/api/v1/ is public and requires no authentication. The official Firebase API at https://hacker-news.firebaseio.com/v0/ is also open. There is a published rate limit on the Algolia endpoints, so cache your results and poll on a fixed schedule rather than continuously. For a founder tracking a handful of keywords, neither limit will be a constraint.

Can I submit my own product to Hacker News?

Yes, through Show HN, provided it follows the Show HN rules: it must be something readers can actually try rather than a waitlist or a demo video, it should be your own work, and you should be in the thread answering questions. Submitting your own company blog posts to the regular pool is also allowed, but do it sparingly and only for posts that stand on their own as technical content. What is not allowed is asking anyone for upvotes, using more than one account, or submitting the same thing repeatedly.

How many times can I mention my product in comments?

There is no numeric rule, and treating it as a quota is the wrong frame. The working test is whether the mention is the direct answer to the question in front of you. If someone asks how to solve a problem your product solves, describing it with a disclosure is fine and often welcome. If you are steering unrelated threads toward it, that is promotion and it will be flagged. Comment histories are public and one page, so anyone can see your ratio in about ten seconds.

Is there a best time to post on Hacker News?

The site-wide folklore is less useful than measurement in your own category. Use the Algolia API to pull a year of stories on your topic that exceeded a points threshold, bucket their created_at_i values by hour and weekday in your audience's timezone, and post where the successful ones cluster. The ranking formula rewards early votes strongly and decays steeply, so what matters most is submitting when the readers who care about your specific topic are awake and browsing /newest.

What happens if my submission gets no attention at all?

Most submissions get none. The /newest page moves quickly and the majority of stories never reach the front page. If you believe a post genuinely deserved a look, the moderators operate a second chance pool that re-surfaces missed submissions, and emailing them about it is legitimate. Reposting the same link soon after is not; the guidelines allow a repost only after a long gap and only if the original got no traction.

Should I automate my Hacker News comments?

No. Automate discovery, never writing. A scheduled job that runs your search queries and hands you a filtered list of threads is a reasonable use of automation, and it is the difference between a routine that survives and one that dies in week three. Generated comments read as generated, violate the spirit and the letter of the guidelines, and put the account at risk. The contribution you are making is your specific first-hand experience, which is exactly the part that cannot be automated.