Data source
Hacker News
Every month the Hacker News whoishiring account opens a thread where companies post jobs, and we count the top-level comments. We also count how many comments mention a fixed list of phrases each month. A sample of comments is labeled for stance by an AI model, Claude Haiku. We publish only totals, never comment text or usernames.
Provides measurements used by the linked signal.
Counts derived from the Hacker News API and the Algolia HN Search API, Y Combinator. Comments belong to their authors and are not stored.
Visit the sourceUsed for: Developer sentimentLicense: No data license published
Monthly threads184
Latest month319
Average, last 12356
Busiest month1041
What the data shows
Computed from every record, as of 2026-10.
- The latest thread (2026-10) has 319 job posts. The busiest was 2021-06, with 1041.
- Threads averaged 356 posts over the last 12 months, against 361 in the 12 months before.
- "ai slop" grew the most, from 6.3 to 17 mentions per 10,000 items between the previous 12 months and the last 12.
- In a sample of 1179 comments mentioning programmer replacement, vibe coding or junior developers, 31 argued that software is getting solved, 338 argued that it is not, and 810 took no clear position. The sample is too small to show a trend.
- Hacker News grew from 3,765,262 to 4,543,430 stories and comments between those periods.
Stance in sampled comments
| Period | Comments | Getting solved | Not solved | Unclear |
|---|---|---|---|---|
| Previous 12 months | 579 | 18 | 176 | 385 |
| Last 12 months | 600 | 13 | 162 | 425 |
| Both periods | 1179 | 31 | 338 | 810 |
Phrases in comments, per 10,000 items
| Phrase | Previous 12 months | Last 12 months | Mentions, last 12 months |
|---|---|---|---|
| "vibe coding" | 10.3 | 14.5 | 6591 |
| "software is solved" | 0 | 0.1 | 23 |
| "replace programmers" | 0.3 | 0.1 | 47 |
| "junior developers" | 1.3 | 0.7 | 335 |
| "layoffs" | 8.2 | 7.4 | 3350 |
| "ai slop" | 6.3 | 17 | 7703 |
Average job posts per thread
| Year | Average posts | Threads |
|---|---|---|
| 2011 | 202 | 8 |
| 2012 | 232 | 12 |
| 2013 | 309 | 12 |
| 2014 | 386 | 12 |
| 2015 | 625 | 10 |
| 2016 | 714 | 12 |
| 2017 | 839 | 12 |
| 2018 | 858 | 12 |
| 2019 | 745 | 12 |
| 2020 | 625 | 12 |
| 2021 | 915 | 12 |
| 2022 | 675 | 12 |
| 2023 | 389 | 12 |
| 2024 | 360 | 12 |
| 2025 | 372 | 12 |
| 2026 | 343 | 10 |
Limitations
- It reflects a startup-heavy, English-speaking audience.
- Top-level comments are mostly job posts, but not all, and a post can list several roles.
- Deleted comments can still be counted.
- A phrase match is not an opinion: a comment can mention a phrase to disagree with it.
- In a sample check, about 85% of "vibe coding" matches contained the exact phrase, so search counts are approximate.
- Stance labels come from an AI model and make mistakes, especially with sarcasm. Every "getting solved" label was checked by a second AI reviewer and 11 of 39 were corrected. The much larger "not solved" group was not fully checked, so it likely contains errors.
- The sample covers up to 50 comments per month that mention specific phrases, so it does not represent all of Hacker News.