7 Hacker News Buying Signals in 2026 (With Real Threads to Search)

Seven patterns on Hacker News reliably indicate that someone is shopping: an "alternative to" request naming an incumbent, a "what do you use for" question, a "how is this different from X" comment under a Show HN, a vendor name sitting beside a price complaint, migration language such as "migrated off", an "anyone using this in production" question, and a "how do you handle" operational pain post. The highest-intent of those are usually the threads with the fewest upvotes, because a thread that never reaches the front page also never gets a reply.

This guide covers each pattern with real threads you can open today, the query syntax that surfaces them, why Hacker News is unusually hard to monitor compared with Reddit, and is honest about where the tooling for this falls short, including ours.

Why a keyword alert doesn't work on Hacker News

Search Hacker News comments for "too expensive" and 3,920 results come back. Read the first page of them. On the day this was written they were about Mac Studio RAM prices, public bathrooms, ocean temperatures, mosquito spraying at airports, and whether a streaming service overspent on fonts. Not one was a buying signal. In the top twenty results, nobody named a tool they currently pay for and asked what to use instead.

That mix churns daily, because results arrive newest first. The character of it does not change.

Now search "Datadog" "too expensive" and 18 results come back, of which roughly twelve are first-person cost complaints or switching intent:

"I love Datadog in general but it's too expensive at this stage."

Comment, 2019, in Ask HN: Which tool do you use to monitor your system and application?

"Is Datadog worth it? I thought it was too expensive, although they do have a lot of features."

Comment, 2021, in Ask HN: Solo-preneurs, how do you DevOps to save time?

"The cost of Datadog is a meme these days."

Comment, 2024, in Saving $10k/Month on Analytics, Snowplow Serverless Alternative

Two hundred times fewer results, and vastly more of them worth reading. The keyword is not the signal. A named vendor next to a named pain is the signal.

This matters more on Hacker News than anywhere else, because the register of the site is uniformly caustic. "This is terrible" and "this is great" are frequently the same commenter about the same tool in the same thread, so sentiment does not separate a buyer from a bystander. Intent does.

The 7 buying signal patterns on Hacker News

1. "Alternative to X" or "instead of X"

A named incumbent plus stated intent to leave it. There is no higher-intent sentence on a public forum, because the budget already exists and the only open question is who replaces the vendor.

Real threads: 

• Ask HN: What should I use instead of Heroku? (82 points, 72 comments).
• Ask HN: What do you run instead of Datadog? (13 points, 21 comments).
• Ask HN: European alternative to Vercel/Cloudflare for hosting (13 points, 20 comments).
• And Ask HN: Alternatives to Hotline Telegram CRM? at three points and zero comments, which is a person publicly asking how to replace their CRM while not one vendor showed up.

2. "What do you use for X"

The category is named and a vendor shortlist is explicitly requested. The comment section becomes the shortlist, which means anyone absent from it loses the deal without knowing it existed.

Ask HN: What do you use for authentication and authorization? is the famous one at 457 points and 236 comments. The useful ones are quieter: Ask HN: What do you use for user management/IAM in your SaaS app? at three points, and Ask HN: What do you use to monitor cron jobs and queue workers? at four points.

3. "How is this different from X" under a Show HN

A prospect performing your competitive positioning in public, and naming the incumbent that owns the default slot in their head. Three examples from August 2026:

"How is this different from descript? It seems like descript, but you took out the video functionality?"

Comment, 17 Aug 2026, on Show HN: Vocal Slice

"How is this different from headroom? Mcptoon seems like it's specific to tool calls."

Comment, 11 Aug 2026, on Show HN: Mcptoon

"How is this different from what's built into Claude?"

Comment, 13 Aug 2026, on Show HN: MCP Memory

4. A vendor name beside a price complaint

Not churn yet, a displacement window. The contract is live, the pain is public, and the renewal conversation is coming. Worth noting where these actually sit: the comment "The cost of Datadog is a meme these days" appears in a thread titled Saving $10k/Month on Analytics, Snowplow Serverless Alternative. Nothing in that title mentions Datadog.

5. Migration language

Churn in flight. A search for "migrated off" returns 303 comment hits and "cancelled our" returns 105. Three of them:

"I was a 1Password user for 18 years and just migrated off it last month" appears in a thread about the Dropbox CEO stepping down. "I have to wonder what happened at Anthropic. We've cancelled our subscription" appears in a thread about Qwen model rankings. "I migrated off Patreon awhile back. It's extremely clunky" appears in a thread about Patreon layoffs.

Read those placements again. Not one appears in a thread about the product being left. Title-level search finds none of them, and watching a competitor's own submissions finds none of them either. Only full-text comment monitoring does.

6. "Anyone using X in production"

Mid-evaluation, and the specific thing being requested is a reference customer. 

• Ask HN: Is anyone using the A2A protocol? (96 points, 45 comments). 
• Ask HN: Anyone using OpenAI's Agent SDK in production? (8 points, 3 comments). The build-versus-buy variant, "self-hosted alternative to", is the same buyer at the same stage weighing your product against a weekend of their own time.

7. "How do you handle X"

Pain articulated before the category is known. Longest cycle of the seven and the largest prize, because whoever names the problem tends to own the frame for the whole evaluation. 

• Ask HN: How do you handle marketing as a solo technical founder? (147 points, 112 comments). 
• Ask HN: How do you handle VAT / Sales Tax accounting as B2C SaaS? (97 points, 86 comments).

One more worth knowing, credited where it is due. The monthly Ask HN: Who is hiring? thread names stacks, team sizes and funding stages in one place, posted by the whoishiring account on the first weekday of every month. August 2026 ran to 237 points and 385 comments. Syften documents the author:whoishiring filter for catching each new thread, and they got there first.

The highest-intent threads are the ones nobody upvoted

Look again at patterns 1 and 2. A CRM replacement request at three points with no replies. An IAM vendor question at three points. A cron monitoring question at four points.

The mechanism behind this is unglamorous. The /ask and /show pages require a minimum score before a submission appears on them, so a thread that does not catch on sits on /newest and then is gone. Nobody saw it, so nobody answered it, so it stayed at three points.

Now consider the arithmetic of being useful. The 457-point authentication thread already holds 236 answers, so a new comment there is number 237, below the fold, competing with people who arrived two days earlier. The three-point thread holds nobody.

Points measure how many people saw a thread. They say nothing about how much intent sits inside it, and on Hacker News those two numbers routinely pull in opposite directions.

Why searching Hacker News manually doesn't scale

The official API has no search in it. The documentation describes it as "essentially a dump of our in-memory data structures", so items are traversed by ID. Searching Hacker News from that API means ingesting and indexing all of it first. There is no authentication and no stated rate limit, which is generous, and also beside the point.

The search index that does cover comments ranks by points, not recency. Comment text is a first-class indexed attribute, as you can see in the open-source index con uration, so comment search genuinely works. The default endpoint sorts by score, which systematically buries the fresh low-score threads that this guide has just spent 400 words identifying as the valuable ones. A by-date endpoint exists and is the one to use.

Boolean OR is not supported by default. A query for "we moved off" OR "migrated off" returns zero results, while "migrated off" alone returns 303. The index treats every token as required, including the literal word OR, so the compound query asks for a comment containing all of it. Every phrasing a buyer might use is therefore a separate request plus a deduplication step.

Quotation marks do not enforce a phrase either. This one catches people out. A search for "expensive too far" returns hits in which that phrase appears nowhere, because the index requires all the words and uses their proximity only to rank the results. So the counts above are word-intersection counts, not phrase counts, and any workflow that assumes phrase matching is quietly looser than it looks.

Scraping the site is not the fallback. The robots.txt sets a 30 second crawl delay and disallows /context?, where comment threading lives. The site FAQ is blunt: request many pages quickly and your IP address might get banned.

There is no notification layer to fall back on. Hacker News has no keyword alerts and no follow-this-topic feature, and does not appear to notify you when someone replies to your own comment. Monitoring is not an upgrade over the built-in alerts, because there are none.

How to monitor Hacker News for these patterns

Start with the cheapest option that fits, because for a small keyword list it is the right answer. F5Bot has a free tier, covers Reddit, Hacker News and Lobste.rs, scans both submissions and comments, and alerts within a few minutes, which is faster than most paid tools including this one. Its constraint is in how matching works rather than in what it covers: no regex, no wildcards, no fuzzy matching, and you cannot type AND, OR or NOT into the keyword field itself. Boolean-style logic does exist on its paid plans, through separate include-all, include-any and exclude flags, and it ships AI filtering as well. So the vendor-plus-pain query is expressible there, using flags rather than operators, on a paid plan.

Syften scans Hacker News using the official Hacker News API and has the deepest filter syntax of the three, with wildcards, site:, type:, author: and natural-language acceptance rules. Its constraint is volume: results are capped per day by plan, at 100 a day on the $29.95 tier and 500 on the $119.95 tier.

Where Boolean search earns its place is pattern 4 and pattern 5, where the useful query is a vendor name intersected with a pain phrase and a set of exclusions:

"datadog" AND ("too expensive" OR "migrated off" OR "alternative to") NOT hiring

Hacker News monitoring here covers stories and comments matching your keywords, on a configurable scan interval per keyword, as often as hourly and defaulting to every two hours. That is slower than F5Bot, and the honest trade is matching precision and per-mention scoring against raw latency. Each match is scored against a short plain-language description of what you actually want surfaced and what to skip, which is what stops the ocean temperature results reaching Slack. If you would rather query it from an agent than watch a feed, there is an MCP server for Hacker News that answers the same questions in natural language.

None of this requires a product, and the mechanism matters more than the tool. The Algolia endpoint plus a scheduled script plus one model call per result will reproduce most of it. The point is that the unit of filtering moves from a word to an intent, which is the same shift that makes Reddit intent monitoring work. For the wider landscape, including tools with agent access, see the best social listening tools in 2026.

FAQ

Can you monitor Hacker News for keywords?

Yes, but not with anything built into the site. Hacker News has no keyword alerts, no email notifications and no topic following, so monitoring requires a third-party tool or your own script against the search API.

Can you search Hacker News comments?

Yes. Comment text is indexed as a first-class attribute in the search index behind the site's own search box, so full-text comment search works. This matters because most buying signals on Hacker News are comments rather than submissions.

Does Hacker News have an API?

Yes, and it has no search. The official API is documented as a dump of in-memory data structures, traversed by item ID, with no authentication and no stated rate limit. Searching requires the separate Algolia-backed search endpoints or your own index.

Is there a free Hacker News monitoring tool?

F5Bot has a free tier that covers Reddit, Hacker News and Lobste.rs with alerts in a few minutes, which is enough for a handful of keywords. The trade is matching precision: no regex, wildcards or fuzzy matching, and Boolean-style logic requires a paid plan and its flag syntax rather than operators in the keyword field.

How do you find leads on Hacker News?

Search comments, not just titles, for the seven patterns above, sort by date rather than points, and reply fast on the low-score threads where nobody else has. The high-score threads are already saturated by the time most people find them. The same approach applied to other platforms is covered under lead generation.

What is the Ask HN Who is hiring thread?

A recurring thread posted by the whoishiring account on the first weekday of each month, in which companies list open roles. Comments routinely name the team's stack, size and funding stage, which makes it a monthly snapshot of who is buying what.