Is Hacker News still a good predictor of tech trends?

In 2021 James Potter asked whether Hacker News predicts tech trends by charting HN title mentions against Google Trends for ten topics. His answer was yes, usually by one to three years. Five years later we re-ran the study with the data through October 2026, and the answer has changed: on the topics of the last four years, the lead is gone.

[Method: we pulled every Hacker News story whose title contains each topic phrase from the Algolia HN API, 106,749 stories across 18 topics, and divided monthly counts by all stories submitted that month, which fixes the HN-traffic-growth problem Potter noted he could not adjust for. Each series is a 3-month average indexed to its own peak. Google Trends series are cut at May 2025, before a mid-2025 rescaling in Trends data, and re-indexed. "Lead" is the gap between the two series first reaching 20 on their own 0-100 index.]

The lead, topic by topic

The pattern sits in the dates, not the topics. Everything that went mainstream before 2020 gave HN readers a long head start. Everything after 2022 reached Google search within a quarter of reaching the HN front page. We read that as the mainstream getting faster, not HN getting slower: ChatGPT reached the general public within a quarter of launch, so there was no window left for a forum to be early in. One honest caveat: we picked topics that did go mainstream, so this list cannot say how often HN is loudly early about things that never arrive.

Bitcoin, still the record

The first mention is from February 3, 2009, a month after the network went live, the same find Potter made. HN attention crossed our threshold in May 2011 and peaked in December 2013. Google search interest crossed in August 2017 and peaked that December. HN had mostly moved on four years before the world's peak interest.

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

3D printing, the clean early call

First mention November 2007, threshold crossed in July 2011, fifteen months before Google. HN interest peaked in June 2013 and never came back; search interest stayed on a high plateau for a decade. Potter's verdict holds unchanged.

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

Kubernetes, covered from day one, dropped five years early

The launch write-up "Kubernetes: Docker cluster management by Google" is from June 10, 2014, the day Google announced it. HN crossed the threshold in March 2016, a year before Google, peaked in January 2019, and has cooled since. Google search peaked in February 2024, five years after HN's peak.

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

Serverless, loved then litigated

HN crossed in May 2016, ten months before Google. HN's peak was October 2018; the single biggest serverless story since is "Serverless: slower and more expensive" with 1,787 points. Search interest kept growing into 2024 while HN argued. On YC company homepages the argument is settled by usage: Vercel is on 27.3% of the 5,386 we crawled in September 2026 and Netlify on 2.9%, counted in our YC group data.

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

Rust, the longest hype that kept being right

Mozilla's language reached HN on launch day, July 8, 2010. The title series crossed in November 2014, five months before search did. The surprise is the right edge of the chart: HN's Rust share peaked in November 2025, fifteen years after first mention and still climbing through 2026. (This series counts titles with the word "Rust" minus "Rust Belt" stories; the videogame is in there too, and it is not what HN front pages.)

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

NFTs, where HN showed up with the backlash

The first NFT story on Hacker News is from 2012 and is about hydroponics, where NFT means nutrient film technique. The first crypto NFT story came in May 2018. HN crossed the threshold in February 2021, one month before Google, and the top NFT story of all time is "NFT projects are just MLMs for tech elites" with 817 points. Potter found HN mentioned NFTs early but quietly; with five more years of data the verdict is harsher: by the time HN talked about NFTs at volume, everyone did. Both curves peaked the same month, January 2022. Last 12 months: 25 NFT stories.

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

Remote work, the miss that stayed a miss

Potter expected earlier HN coverage than he found. Same here: Google search crossed the threshold in December 2013, HN in July 2015, and both peaked in April 2020 for the obvious reason. Working remotely was a mainstream topic first and a front-page topic second.

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

ChatGPT and the end of the head start

HN crossed the threshold in the launch month, November 2022; Google search followed in February 2023. Three months of lead. Stable Diffusion in 2022: two months. Vibe coding, coined in February 2025, went from first HN story to measurable Google volume in a month. The Model Context Protocol launch post (872 points, November 2024) and the first MCP search volume are a few weeks apart; on series this young the ordering is noise, so we call it simultaneous. AI agents: Google crossed two months before HN did.

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

So is HN still a leading indicator?

Split the question in two.

HN did not get slower. On every topic in the table it covered the thing at or near launch: Kubernetes on announcement day, ChatGPT in launch week, MCP on launch day. What changed is the other curve. Google search needed six years to catch up on Bitcoin and one on Kubernetes; it needed three months on ChatGPT and one on vibe coding. The head start came from the rest of the world being slow to notice, and on AI topics the world now notices within weeks, so there is no spread left to trade on.

One set of series does point the other way. On AI agents, Google search crossed two months before HN did, and the biggest AI-agents story of all, "An AI agent published a hit piece on me" with 2,346 points, is backlash coverage, like the NFT top story before it. Backlash follows the mainstream by definition, so on AI topics HN reads less like a scout and more like a critic.

Where the old pattern survives is lower on the stack, in the things only developers care about so far. Three series look today the way Kubernetes looked in 2015:

• Local-first. The 2019 essay started it; 663 of its 940 all-time HN stories came in the last 12 months, and the series peaked in May 2026. Google search volume for "local-first software" in the year to May 2025: zero.
• Self-hosted. Crossed on HN in June 2015, 46 months before Google, and its HN share peaked in November 2025, ten years in and still rising.
• Zig. First language story February 2016, crossed on HN in November 2020, twenty months before search, HN peak November 2025.

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

Sources: Algolia Hacker News API (October 2026); Google Trends (cut May 2025)

RISC-V ran the same shape a cycle earlier (HN crossed twelve months before Google). Not everything HN loves goes this way: e-ink and homelab both reached Google before they reached the HN threshold. The list above is where the gap is open right now.

How we counted

HN stories from the Algolia Hacker News API, October 14, 2026: every story whose title contains the topic phrase, case-insensitive, 106,749 stories over 18 topics. Monthly counts are divided by all stories submitted that month (10.08 million across 2007-2026), smoothed with a 3-month centered average and indexed to the topic's own peak. Rust excludes "Rust Belt" titles, Zig excludes Zig Ziglar, and MCP counts start November 2024 because the abbreviation meant other things before. Google Trends series were pulled per term (timeframe "all"), cut at May 2025 because Trends data rescales mid-2025, and re-indexed to the cut window; "local-first software" had no measurable volume in the final year, so its chart has no Trends line. Threshold crossings compare each series against 20 on its own 0-100 scale, with HN crossings requiring at least 3 stories a month. Points totals are values at pull time. The YC homepage counts come from our crawl of 6,133 Y Combinator company sites in September 2026, 5,386 usable, the same crawl behind HN hype vs. what YC startups install.

Dataset

Monthly series for all 18 topics (HN counts, HN totals, Google Trends), the per-topic summary with first-mention links, and the full 106,749-story pull are in static/data/hacker-news-lead-over-google, CC BY 4.0.