Recomputed every morning · 2026-10-20
How long does a role stay listed?
Every job board shows you a posting date and leaves you to guess what it means. This measures the one thing that can be measured from the outside: how many days 378 roles stayed on their employer's careers page before disappearing from it.
62 days
The median time from the date an employer published a role to the date it disappeared from their careers page. A quarter were gone within 24 days, a quarter lasted longer than 137. Counting only the 378 removals this board watched happen, between 2026-08-25 and 2026-10-20, having started checking on 2026-08-22.
What this does and does not tell you
A posting is not a deadline, and the spread here matters more than the median. 16% of these roles were gone inside two weeks and 37% were still listed after three months, so how long a listing has been up says very little about whether it is still live.
What is being counted is disappearance from the employer's own careers feed, checked daily. That can mean the role was filled, that it was withdrawn, that headcount moved, or that the company reorganized its job board. This board cannot see which, and nothing here should be read as a hiring statistic, a confirmed hire, or a deadline. These are also aggregate figures over roles that have already closed; they do not predict what any one listing still open will do.
Alignment & interpretability listings come down soonest
Median days listed by discipline, for categories with at least fifteen removed roles behind them. A shorter median means those listings tend to disappear sooner, not that those roles are filled faster.
| Discipline | Median days listed | Roles measured |
|---|---|---|
| Alignment & interpretability | 31 | 23 |
| Red team & safeguards | 50 | 31 |
| Evals & benchmarks | 56 | 139 |
| Human data | 70 | 124 |
| Post-training & RL | 116 | 54 |
The shortest-lived postings
The postings that spent the fewest days on their employer's careers page before disappearing, counting only roles that name the discipline in the title. Why each came down this quickly is not visible from outside, and a short listing life is as consistent with a withdrawn requisition as with a fast hire.
| Role | Company | Days listed |
|---|---|---|
| Associate, ML Data Operations, GO-AI Operations | Amazon | 1 |
| AI Model Policy Trainer, Generalist (Remote) | Handshake AI | 1 |
| Head of Reinforcement Learning Operations | Hippocratic AI | 2 |
| AI Safety Policy Evaluator, Violence & Threats | Remote - US Only | Handshake AI | 2 |
| Senior Offensive Security Engineer (Red Team) | Salesforce | 2 |
| Offensive Security Engineer - Red Team | Salesforce | 3 |
The longest-running before they came down
The postings that stayed on their employer's careers page longest before disappearing. A long listing life is not evidence that a role was hard to fill: employers leave evergreen requisitions up, repost the same job, and sometimes forget to take one down. This board can see the duration and not the reason.
| Role | Company | Days listed |
|---|---|---|
| Staff+ Software Engineer, Safeguards ML Infrastructure | Anthropic | 438 |
| Senior Software Engineer, ML Evaluation Infra and Efficiency | Waymo | 415 |
| Staff Machine Learning Research Scientist Agent Post-training - Enterprise GenAI | Scale AI | 320 |
| Machine Learning Systems Research Engineer/Scientist, Agent Post-training - Enterprise GenAI | Scale AI | 320 |
| Software Engineer, Quantitative Evaluations | Waymo | 284 |
| Software Quality Operations Specialist, Safety Evaluation | Waymo | 279 |
Free. One email on Thursdays.