Historical Hacker News hiring advertisements
HN hiring posts: volume, timing and changing AI language.
Complete comparisons through June 2026.
The analyzed posting stream contained 3,786 eligible posts in 2025. That is 61.3% fewer than the observed 2021 peak of 9,780. The latest matched half-year changed by -1.2%.
Between those same complete-year endpoints, posts with detected AI-related language increased 18.7%. A rising language share and a growing posting stream are different findings.
These counts describe analyzed hiring advertisements on one website. A post can describe several roles or repeat an earlier advertisement. They do not count vacancies, unique employers, hires or the whole startup labor market.
The long-run size of the posting stream
A higher share for one kind of advertisement can coexist with fewer advertisements overall. Start with the size of the source before interpreting a percentage. The annual view includes only complete, non-provisional years.
Shared scale: zero to 9,780 posts. Cyan marks the latest complete year.
Each point is one monthly count. Hollow amber points are provisional. Missing months have no point; they are not recorded zeros. These counts are not seasonally adjusted.
Observed monthly records: Apr 2011–Aug 2026. The last non-provisional month is June 2026. Exact values are available below and in the download.
Compare the same six months
January–June 2026 contained 1,831 analyzed eligible posts, compared with 1,854 in January–June 2025. The difference is -1.2%. Both windows contain six complete, non-provisional months. A full year would be the wrong baseline for this comparison.
January–June 2025 · Jan 2025–Jun 2025
January–June 2026 · Jan 2026–Jun 2026
The 2021 annual peak is useful historical context, not a neutral baseline for the latest period. Changing participation, posting conventions and where companies recruit can all affect the observed stream. These data do not identify the cause.
AI language and total posting volume can move differently
The observed peak and latest complete-year counts put language shares in the context of the source’s total size. The green segment counts posts with detected AI-related language. The gray segment is the remaining eligible posts, without detected AI language.
Shared scale: zero to 9,780 posts. Segment counts sum to each year's eligible total.
1,588 with detected AI language · 8,192 without detected AI language
1,885 with detected AI language · 1,901 without detected AI language
Between these two years, the count with detected AI language changed by +18.7%. The complementary count changed by -76.8%. Read the counts together before interpreting the share.
A language mention does not identify an AI occupation, prove AI displaced a job, or show which tools a team uses. Detection may miss terms and its sensitivity can differ across eras. The AI Hiring Reality Report explains the language measure in more detail.
Calendar patterns depend on which years you compare
Dividing each month's count by its own year's total puts large and small years on the same footing. A month in an evenly distributed year would account for 8.33%. The heatmap shows each year's actual shares. Partial and provisional years are excluded.
One color scale across every year: below 7.5%, 7.5%–9.5%, above 9.5%. The exact percentages carry the result; darker or lighter cells alone do not establish a trend.
The line marks the median; each band contains the middle half of year-specific shares. This is variation across years, not a confidence interval. Every row uses the same scale.
Cyan reference: 8.33%
A quiet January can reflect the year's direction
Annual normalization does not remove growth or decline inside a year. To check that problem, split each complete year by whether July–December contained more, fewer or exactly as many posts as January–June. This is a descriptive grouping, not a seasonal adjustment or a causal test.
Read this check beside the pooled calendar profile. If January's pattern differs between the two directions, the pooled result cannot identify a dependable January lull. Even the grouping can contain genuine calendar effects. Separating those from trend would need a different model and evidence.
Month length, holidays, thread schedules and changes in who uses HN are not controlled here. ONS explains why seasonal adjustment separates calendar variation from longer-term movement. Our charts remain descriptive.