Sign in

Method

How Glassbox counts

Every number on this site comes from one count of live jobs. This is how it's made, and what it leaves out.

01

Where the jobs come from

Every day at 05:00 UTC we read public job boards.

Many jobs come from employers' own boards: 181 companies on Greenhouse, Ashby, Lever and Workable. We chose that list by hand.

The rest come from job sites: RemoteOK, Remotive, Arbeitnow, Jobicy, Himalayas and We Work Remotely, plus six Kenyan job feeds.

Jobs found during someone's own CV scan, and links people paste in, are never counted. They are one person's search, not the market.

02

Which way the list leans

The hand picked list leans toward US employers, AI companies and fast growing tech firms. Many of their jobs are senior.

So our count reads more senior than the wider job market. When we checked in October 2026, about 2 in 100 of our tech jobs that state years asked for a year or less. Of US tech jobs that state years, Indeed counted 18 in 100 in 2025.

Big employers that post on their own career sites, or on Workday, aren't in the count yet.

03

One job, counted once

A job counts while it is live. A company board or job site must have listed it in the last 3 days. A feed, in the last 30 days. Nothing posted more than 90 days ago counts.

The same title at the same employer counts once, however many boards or cities list it. Two separate openings with the same title at one employer also count once.

04

How skills are read

Skills in jobs are read by fixed rules, not by a language model. We keep a list of 250 skills, each with the ways jobs spell it, and look for them in the job's text as whole words.

A skill under a heading like Requirements counts as required. Under a heading like Nice to have, or softened with words like ideally or a plus, it counts as optional. A skill in the job title counts as required.

The employer's own name is never counted as a skill.

A job under 800 characters is too short to read. It counts as a job, but not toward skills.

Your CV is read another way: a language model reads the skills on it. What happens to your CV explains where it goes.

05

How job types, levels and years are read

The job type, like backend or data science, comes from the title. For AI and machine learning titles, the text decides whether the work is training models or building apps.

The level comes from the title first: words like intern, graduate, junior, senior, staff or principal. Then the board's own level, if it gives one. Then wording senior jobs use, like "proven track record". Then the years the job asks for.

Years are the highest figure a job requires. A range counts from its lower end. "Ideally 5 years" is optional, not required. "No experience required" counts as zero.

When a rule changes, every job is read again with the new rule.

06

Entry level, on the entry level page

A job counts as entry level if its level reads as intern, entry level or junior. Or if its title says intern, graduate, new grad, entry level, early career or junior. Or if it requires 2 years of experience or less and its level isn't senior.

Software means five job types: software engineering, backend, frontend, full stack and mobile.

A page shows shares only when at least 100 of its jobs are long enough to read. Below that, a share describes a few employers more than the market.

07

How fresh it is

Jobs are read once a day. The counts are rebuilt from them about every hour, and pages show the new count within five minutes.

Every count says the day it was taken.

08

What we can't see

Jobs on boards we don't read, and jobs never posted in public.

Skills a job names in words that aren't on our list.

Whether a job is filled, how many people apply, or who gets hired.

Back to the count