Measure how long jobs stay open at tech companies
An agent read 25 open jobs each at five companies and measured how long each had been posted, e.g. 21 of 25 at Figma were 60+ days old.
Done with a custom-built or unnamed agent
- What they did
- The agent used a tool called Company Jobs Finder. It found each company's Greenhouse board, read the public API's first-published date for the first 25 listed jobs at Stripe, Airbnb, Figma, Databricks and Anthropic, and computed each posting's age. It needs no key. Filters for keyword, location and remote work are available.
- How it went
- Figma had 21 of 25 postings at 60+ days, with a median age of 269 days. Stripe had 7, Airbnb 8, Databricks 13 and Anthropic 9. Databricks had the oldest posting, at 1,031 days.
- Worth knowing
- The sample is only the first 25 jobs in each board's own order, and reposts reset a posting's age, so old postings don't prove ghost jobs.
Try it yourself with Claude
For each of these companies: [company names], look at their careers pages and sample about [25] open roles. Find out how long each role has been posted and report what share are older than [60] days. Give me a table per company and one sentence on what the numbers suggest. Stop once every company has a sample, and note any company where posting dates were not available.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: tools.openkrill.app · Oct 11, 2026 · Submitted by OpenKrill
Related use cases
Find connections across 100 non-fiction books
Claude Code browses a personal library with CLI tools and surfaces trails of related excerpts across books.
Collate WHO data into a multi-tab spreadsheet
Testers had ChatGPT Agent navigate the WHO site, extract the latest data and organise it by indicator, country and trend; it often fails on logins or dynamic pages.
Compile recent LLM research papers into a Word doc
Comet found major LLM/agent training papers from six months with links and analysis, then saved them to a Word doc via Outlook.
Grade Cursor on a real data science project
A data scientist tested Cursor on a real project: weak on pandas, data cleansing and scraping, strong at prototyping.
Run a systematic literature review from one prompt
SciSpace's agent generates PICO questions, searches and screens papers, downloads PDFs, extracts data and outputs the full review with a PRISMA diagram.
Custom agent
Give a coding agent a paper corpus to find new techniques
Poster gave a Claude Code agent access to about 2M papers to search before each attempt; it found techniques it otherwise wouldn't know and beat a baseline agent.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.