Three days, three different top fives
League tables of employers are usually dull because they barely move. This one moved twice in 48 hours. Okta and Block have dropped out entirely; Databricks left on the 7th and returned on the 8th; and two names that were nowhere on the 6th are now second and fourth.
The direction is as interesting as the movement. Rippling went from 73 live roles to 114 in one day, and Anthropic from 72 to 98 — while Stripe, still the largest, fell from 174 to 123 across the same window. The biggest hirer in the aggregate is contracting while the fastest climbers are an AI lab and an HR platform.
| Employer | 6 Aug | 7 Aug | 8 Aug | Two-day change |
|---|---|---|---|---|
| Stripe | 174 | 155 | 123 | −29% |
| Rippling | — | 73 | 114 | +56% in a day |
| Datadog | 130 | 91 | 102 | −22% |
| Anthropic | — | 72 | 98 | +36% in a day |
| Databricks | 85 | — | 91 | +7% |
| Okta | 82 | 70 | — | out of the top five |
| Block | 77 | — | — | out of the top five |
Why we will trend this, having refused to trend salaries
Yesterday this desk retired day-over-day salary comparisons, after publishing a British finding that reversed itself in twenty-four hours. It is fair to ask why employer counts get treated differently.
Because they are a different kind of number. An advertised median is computed over only the listings that disclose pay — a small, self-selected subsample, so a handful of new ads can swing it wildly. An employer's live count is a census: every listing we hold for that company, counted. Nothing is inferred and nothing is sampled. A census of hundreds moves when the world moves; a median of a dozen moves when the dozen changes. That is the whole distinction, and it is why one is publishable as a trend and the other is not.
The same shape, at global scale
What we are seeing in five rows is the local edge of something much larger. AI job postings finished 2025 running 134% above pre-pandemic levels while overall tech postings sat 34% below them. The share of tech postings naming at least one AI skill went from 15% in January 2024 to 73% in May 2026, and roles with AI in the title rose 173% year on year in the first quarter of 2026. Analysts have taken to calling it two parallel realities.
The two climbers fit that story precisely. Anthropic carried 1,095 active job postings in 2026, more than double the year before, and is among the most aggressive net hirers in the industry. Rippling has grown to roughly 5,000 staff from 3,600 in 2024, announced 150 new Dublin roles to build out its EMEA base and is doubling its India headcount toward 2,000. Neither is a surprise once you look up from the table.
The pool grew. The remote share did not.
The aggregate itself expanded: 3,864 open roles on the 6th, 4,139 today — up 7.1% in two days. Growth on that scale in that time is unusual and worth watching rather than explaining.
What grew alongside it is telling. The share of listings flagged remote-eligible fell from 15% to 12% while the pool added 275 roles. The arithmetic only works one way: the roles arriving are more office-bound than the roles already there. For a job seeker filtering on remote, the market got bigger and slightly worse at the same time.
| 6 Aug | 7 Aug | 8 Aug | |
|---|---|---|---|
| Open roles | 3,864 | 3,927 | 4,139 |
| Remote-flagged | 15% | 15% | 12% |
| Job sources | 38 | 39 | 38 |
What to take from it
If you are searching: the fastest-growing employers in a market are rarely the most famous ones, and a league table two days old is already wrong. Watch the climbers, not the leader.
If you are hiring: you are competing for engineers against companies that added forty roles in a day. Speed of process is now part of your offer, whether or not you intended it to be.
Every market figure here is a page you can open. Start with the market by city and read your own.
Earlier editions
- · 8 min readA law changed who counts as unemployed in France.French unemployment reached 8.3% in the second quarter of 2026, its highest reading since the pandemic quarter of 2020. INSEE publishes something most coverage leaves out: a decomposition showing that people brought into the count by the 2023 full-employment law account for nearly half of the rise across six quarters. That is not the same as saying the rise is not real - the majority of it comes from a population the law never touched - and holding both facts at once is the whole of this piece.Read the analysis →
- · 7 min readEurope’s unmet need for work is twice its unemployment.Eurostat counts 23.7 million people in the European Union with an unmet need for employment. 12.0 million of them are unemployed. The other 11.7 million are working part-time and want more hours, or want work and are not searching, or are searching and cannot start yet - and the unemployment rate, 5.7%, sees none of them. This is the third edition in a row about a number that leaves people out, and the first about the number built to count them.Read the analysis →
- · 7 min readOne unemployed American per job opening. That is not full employment.The Bureau of Labor Statistics counts 1.0 unemployed person per job opening in the United States - a ratio that has sat between 1.0 and 1.1 for eight straight readings. Yesterday this desk reported the British equivalent at 2.5. The gap is real. What the American number does not say is that anyone who wants a job can have one, and the same release explains why: in the very month the ratio read one-to-one, 7.4 million openings sat unfilled while employers hired 5.3 million people and let 1.8 million go.Read the analysis →