Trends over time
Each snapshot is dated and kept, so the same measures can be tracked as the months go by.
A scheduled job collects a fresh snapshot every Monday and commits it, so the archive builds itself. There are 10 so far (2026-06-17, 2026-06-22, 2026-06-29, 2026-07-06, 2026-07-13, 2026-07-20, 2026-07-26, 2026-07-27, 2026-08-03, 2026-08-10).
| Snapshot | Orgs | Vacancies | EOIs set aside | Early-career | Senior |
|---|---|---|---|---|---|
| 2026-06-17 | 8 | 59 | 10 | 3% | 48% |
| 2026-06-22 | 8 | 59 | 10 | 3% | 48% |
| 2026-06-29 | 12 | 61 | 12 | 8% | 38% |
| 2026-07-06 | 12 | 67 | 12 | 8% | 39% |
| 2026-07-13 | 12 | 63 | 12 | 8% | 40% |
| 2026-07-20 | 12 | 63 | 12 | 8% | 40% |
| 2026-07-26 | 12 | 63 | 11 | 5% | 46% |
| 2026-07-27 | 12 | 63 | 11 | 5% | 46% |
| 2026-08-03 | 12 | 71 | 11 | 6% | 48% |
| 2026-08-10 | 12 | 75 | 12 | 5% | 47% |
Reading these points. Three caveats. Coverage grew from
8 orgs to 12 as more orgs were added, so it becomes a like-for-like
comparison only from the 2026-06-29 snapshot on. The classification rules were revised on
2026-07-26 and every snapshot above is re-scored with the current rules, so the columns agree
with each other but not with figures this site showed before that date. And the orgs collected
by hand only change when someone re-captures them, so their rows carry forward unchanged between
captures; a flat line for those orgs means no fresh reading, not necessarily no change.
The 2026-07-26 step is a correction, not a movement. Early-career
falls from 8% to 5% at the last point, and that is almost entirely one thing: three GovAI winter
fellowships had closed, but because GovAI is captured by hand they kept being carried forward
into the scheduled snapshots until the supplement was re-read on 2026-07-26. The underlying
change happened at some point across those weeks, not on that day. Read the step as the archive
catching up with reality.