Someone indexed 18 years of HN comments. Here's what the data shows.

What happened

A Show HN post titled "Google Trends for Hacker News" hit 345 points, surfacing a new tool at hackernewstrends.com that indexes roughly 18 years of HN comments — not just submission titles — and exposes a time-series graph of how often any term gets mentioned. Type in `kubernetes`, `rust`, `mongodb`, `webassembly`, or `oauth` and you get a chart of mention frequency across the entire archive, with the ability to overlay multiple terms.

The novelty isn't the dataset. The HN corpus has been publicly available via Firebase and BigQuery for years; people have run one-off queries on it forever. The novelty is treating the comment stream — not the front page — as the primary signal. Submission titles tell you what got upvoted. Comments tell you what 800,000 working engineers actually argued about at 11pm on a Tuesday.

The interface is deliberately minimal: a search box, a chart, and overlays. No login wall, no API key gate, no "book a demo." It's the kind of weekend-build tool that exists because the author wanted to answer a specific question and figured everyone else might too.

Why it matters

Most "developer trends" data is downstream of marketing. The Stack Overflow Developer Survey is self-selected and annual. Google Trends measures search intent, which conflates "I'm learning this" with "I'm debugging this at 3am and furious." GitHub stars are gameable and lag adoption by years. The JetBrains and JetBrains-adjacent surveys ask people what they *use*, which is a different question from what they're *thinking about replacing*.

HN comments are a leaky but uniquely honest signal: they capture the moment a senior engineer publicly defends, attacks, or switches off a technology. A spike in `mongodb` mentions in 2011 isn't the same as a 2024 spike — context shifts — but the shape of the curve, especially the inflection points, tends to correlate with real architectural shifts that other instruments miss by six to eighteen months.

A few patterns the tool exposes that are worth checking yourself:

- The `docker` curve peaks around 2016-2017 and then descends, not because Docker died but because it stopped being a topic of argument and became infrastructure. Boring is the absorbing state of successful technology. - `kubernetes` follows the inverse — a long, grinding climb with no clear peak, because the operational complexity keeps generating new things to argue about. - `rust` shows a steep, mostly-monotonic climb starting around 2015, with no "hype peak and crash" shape. That's a different signal than the one you get from Twitter, where every language looks like a fad. - `oauth` shows two distinct waves — the original 2010-2012 wave, and a second, larger wave starting in 2023 driven by the Cloudflare/Auth0/MCP-era authorization conversations.

The community reaction in the Show HN thread is itself a data point: top comments aren't praising the chart, they're posting screenshots of unexpected curves and arguing about what they mean. That's the actual product. The chart is a conversation starter for engineers who already have priors about what the data *should* show.

What the tool *can't* do, and what the author was honest about, is distinguish sentiment from volume. A spike in `electron` mentions could be people adopting it or people complaining about its memory footprint — usually both. Sentiment analysis on a corpus this opinionated and sarcastic is genuinely hard; the naive approaches all fail on HN's specific register. Anyone who claims they've solved it is selling something.

What this means for your stack

Three concrete uses for a tool like this if you're making technology bets:

Timing migrations. If you're evaluating a database, framework, or runtime, plot it against the incumbent. You're not looking for which line is higher — that's a popularity contest. You're looking for the inflection point where the challenger's slope steepens and the incumbent's flattens. That's the moment when senior engineers in the wild started taking the migration seriously. For most technologies, that inflection point precedes the "safe to adopt" point by 12-24 months and follows the "early and risky" point by about the same.

Sanity-checking vendor pitches. A sales engineer tells you their category is exploding. Pull up the term in the tool. If the curve is flat or descending while the pitch deck shows a hockey stick, you've learned something — either the category is real but the term has shifted (check synonyms), or the hockey stick is fictional. Either way, you ask a sharper question on the next call.

Justifying internal proposals. "We should adopt X" lands differently when you have a chart showing X's mention curve overtaking the thing you're currently running, with the inflection point dated to a specific architectural shift. It's not proof — HN is a sample, and a weird one — but it's evidence, and most internal architecture debates run on much less.

The failure mode to avoid is treating the chart as truth. HN's commenter base is disproportionately American, disproportionately Bay Area, disproportionately startup-employed, and disproportionately allergic to enterprise Java. If you're making decisions for a 5,000-person bank in Frankfurt, the absence of `cobol` mentions doesn't mean COBOL is gone. It means the people running COBOL aren't on HN.

Looking ahead

The interesting follow-up isn't a fancier HN trends tool — it's the same treatment applied to other technical comment corpora: Lobsters, the relevant subreddits, dev.to, the GitHub issue trackers of major frameworks. Each has its own bias, but the intersection of biases is where signal lives. The first person to ship a cross-corpus version with honest sentiment scoring — even crude polarity — will have built something genuinely new, and considerably more useful than another year of the Stack Overflow survey.