Show HN Posts Dataset, Categorized by Subject

What’s in the file

26 columns over 68,025 rows. Distinct counts and null rates are measured from the file itself, not estimated.

Things worth noticing

• Agent tooling went from a niche to a fifth of Show HN. Posts about coding agents, agent infrastructure or agent skills were 6.9% of submissions across the first four quarters and 18.3% across the last three. Coding agents alone went 3.6% → 11.3%, and the narrowest slice — skills, prompt packs, CLAUDE.md and AGENTS.md conventions — went 0.5% → 2.4%, a category that barely existed in 2024.
• Claude Code is named in 2,273 submissions. It first appears in February 2025, crosses 1% of all Show HN posts that June, and peaks at 6.8% of everything submitted in March 2026 — roughly one Show HN in fifteen naming a single tool. MCP is named in 2,156, Cursor in 1,007, Codex in 846.
• Volume quadrupled and attention collapsed with it. September 2024 saw 1,531 submissions averaging 19.2 points; February 2026 saw 6,218 averaging 6.4. Around a third of posts never clear a single point in either period, so this is not more failures — it is far more posts chasing a fixed amount of attention, and the GitHub-hosted share rising from 26.1% to 37.1% alongside it.
• The AI posts are the ones being ignored. Across the last three quarters AI-themed submissions are 43.0% of posts but only 36.1% of points — 6.9 points on average against 9.2 for everything else. Vertical AI assistants do worst of all at 2.8. What got squeezed out of the feed is unglamorous: Web & browser fell 5.8% → 3.4%, Education & learning 5.3% → 3.1%, and RAG, the defining 2023 pattern, 1.7% → 1.2%.

Put an agent on it

Anything you were about to do with this file by hand is a job that runs every week forever. Each of these is a working prompt — open it and it is already loaded.

Every Monday, pull the last seven days of Show HN submissions and classify each one by subject and AI theme using the same definitions as the Show HN subjects dataset.

Report:
- Total submissions this week against the trailing 13-week average, as a count and a percentage
- The three subjects and three AI themes that gained the most share, and the three that lost the most, each with both percentages
- Any theme that appeared this week and did not exist in the trailing quarter
- Median points and the share of posts that finished at one point or less, against the same trailing window
- The five highest-scoring posts, with their subject and AI theme

Six sentences, numbers inline, biggest move first. If the week is unremarkable, say so in one line rather than inventing a narrative.

I will name a tool or pattern, for example "MCP", "Claude Code" or "RAG".

From this dataset, build its adoption curve: share of all Show HN submissions naming it, by month, from first appearance to now. Then:
- Identify the month it crossed 1% of submissions and the month it crossed 5%, if it has
- Compare its first 12 months against the first 12 months of two or three comparable names, aligned to each one's own first appearance rather than to calendar time
- Break the current share down by subject, so I can see whether it has spread beyond developer tools
- Report average points for posts naming it against the all-post average for the same months, and say whether attention is keeping pace with volume
- State whether the trailing six months are rising, flat or falling, with the numbers

Finish with three sentences on whether this looks like early adoption, a plateau or a decline. Say explicitly where Show HN's self-selection is likely to overstate the case.

Watch Show HN for submissions in these categories: [name the subjects and AI themes you care about].

Each week, report:
- Every new submission matching those categories, with title, link, points and comment count
- Which ones name a tool from the same stack as ours: [list your stack]
- Any submitter who has now posted three or more times in these categories, and what the arc of their posts suggests they are building toward
- Whether the category's share of Show HN is rising or falling against the trailing quarter

Flag separately anything that clears 100 points — at the current median that is roughly the top 2% of submissions, and it is the signal worth reading the comments on. Skip the rest.

Load it yourself

The file is a plain parquet. Nothing here is specific to us.

-- No download step: DuckDB reads the parquet over HTTP range requests.
INSTALL httpfs;
LOAD httpfs;

CREATE VIEW show_hn_subjects AS
SELECT * FROM read_parquet('https://cotera.co/datasets/show-hn-subjects.parquet');

SELECT * FROM show_hn_subjects ORDER BY story_id LIMIT 20;

More datasets