He indexed 45M Hacker News comments to chart 18 years of tech hype

A solo dev turned 18 years of Hacker News into a Google-Trends-style chart

The pitch on hackernewstrends.com is simple: type a term, get a line showing how often it appeared in Hacker News discussion over time. Type several and they race each other. The site runs a live date-histogram across 45 million posts and comments spanning 18 years, and under each line it shows the actual stories and comments from that window, filterable by term or author.

The maker, posting in the thread, framed it as a weekend toy that got out of hand:

"This was a small project of mine after I've found out that I can simply … the whole hackernews archive (~48GB) and play around with it. … I like that you can discover what went crazy in the timeline, they just come up as small burst of activity."

The bursts are the fun part. lk-99, the room-temperature-superconductor that wasn't, is a single violent spike in mid-2023. Most security stories behave the same way: one peak at disclosure, then silence. The exception, one security engineer pointed out, is Stuxnet: "much more interesting than most other attacks since it was very political and openly published… the thing that attack was about is still a news headline today." There is also a separate Who's Hiring page charting programming-language demand across every monthly hiring thread HN has ever run.

"This is about published text, not searches"

The name sets an expectation the data can't meet, and the most-upvoted correction in the thread came from a long-tenured HN regular:

"Google Trends is about searches. This is about published text. More like if Google Trends counted word occurrences on webpages… People don't write much about non-newsworthy things whereas many people search 'burger' anytime they want a burger delivery. The datasets aren't usable in the same way."

It's closer to Google's Ngram Viewer pointed at a forum than to Trends, a distinction several readers made independently.

The second asterisk is term disambiguation, the exact problem Google spent years solving. The site matches raw strings, so common words and symbols leak. One reader testing editors found whole languages missing: "no Java or Go for the languages? Seems odd… I wondered if 'go' got filtered out because it's also just a regular word." A 14k-karma developer caught a subtler one: "C# actually just matches C in the graph even though it only highlights C# in the example article titles." And another reader realized the buoyant "Grunt" line wasn't the task runner at all, just people writing "grunt work."

The most-repeated feature request is normalization. As one commenter put it, a search for iPhone dips around 2025 — "but did interest in it decline or were there just fewer comments on Hacker News that year?" Without dividing by total HN volume, every line partly tracks the site's own growth, not the topic's.

ClickHouse's maintainer: "you can build this in a single SQL query"

The reaction that reframed the project came from the lead developer of ClickHouse, who pointed out the heavy lifting is already a public utility:

"I host a publicly open database with Hacker News data… So you can create any sort of similar services in a single SQL query and an HTML page. … It is also updated in real-time."

The site itself isn't bulletproof under load — it spent the morning throwing FUNCTION_INVOCATION_TIMEOUT and Upstash rate-limit errors as the front page hammered it. And the auto-generated narratives it writes over the charts drew their own pushback: when the page declared REST gave way to a gRPC/GraphQL split, a developer reading the same graph flatly disagreed — "REST is the default until 2017, GraphQL is briefly popular around early 2020s, then the web returns to REST."

What survives the nitpicks is the relay race only a forum's own text can show. Crypto cresting as chatgpt erupts; Flash bleeding into HTML5; the long slow rise of Rust. The data won't tell you why HN cared — but it timestamps, to the month, exactly when it stopped.