95 listings said remote on the location line and named office days in the text; the index now calls them hybrid

A listing has two places to say where the work happens: the location line, which boards and aggregators read, and the description, which the candidate reads. On 6 November 2026 the index read the full text of every live listing whose location line it had classified as remote, 2,326 of them, and looked for a stated office schedule. 95 listings (4.1%) carried one. 83 stated a number of days or a list of weekdays; 12 used the words anchor days or in-office days without a number. The other 2,231 (95.9%) said nothing about an office. Remote, as the corpus states it, is mostly remote.

Nine employers, five schedules

The 95 came from nine employers, all on Greenhouse- or Ashby-hosted career pages: Asana 35, Brex 14, Vercel 14 (12 with a weekday list, 2 with anchor-day wording), Lyft 10, Notion 10, Watershed 8, Okta 2, Dashlane 1, Coinbase 1. The sentences are the employers' own, as published on the listing pages on 10 November:

Asana, location line "San Francisco (Remote)" and similar: "This role is based in our San Francisco, New York, or Chicago office with an office-centric hybrid schedule. The standard in-office days are Monday, Tuesday, and Thursday."

Brex: "We currently require a minimum of three coordinated days in the office per week, Monday, Wednesday and Thursday."

Vercel: "If you're based within a pre-determined commuting distance of one of our offices (SF, NY, London, or Berlin), the role includes in-office anchor days on Monday, Tuesday, and Friday."

Lyft: "Team Members will be expected to work in the office 3 days per week on Mondays, Wednesdays, and Thursdays. Lyft considers working in the office at least 3 days per week to be an essential function of this hybrid role."

Notion, location line "San Francisco, California (Remote)": "We work from our offices on Mondays, Tuesdays and Thursdays (our Anchor Days) because we do our best thinking and building together in person."

None of these sentences hides anything; each states the schedule plainly. What they show is that the location line and the text can disagree, and that an index reading only the location line will count a three-day office week as remote.

What changed in the index on 9 November

Until 8 November a listing whose location line read remote was classified remote whatever office schedule its text stated. From 9 November (classification method 3, noted with its date on the how-we-score page) a listing whose location line reads remote but whose text states office days is classified hybrid: with a Hybrid Transparency Score when the number of days can be read, with a days-disclosed value of zero when the text says anchor days without a number. Everything else stays remote; nothing was reclassified retroactively. Existing listings change on their next ingest, and the first day shows the size of the step: on 9 November the live index held 2,174 remote, 638 hybrid and 36 unclear listings out of 2,848; on 10 November, 2,164 remote, 779 hybrid and 56 unclear out of 2,999. The hybrid count rose by 141 while the remote count stayed flat as 151 listings entered. Among the named employers the change is visible per company: of Asana's 87 live listings, 82 are now hybrid and 4 remote; of Brex's 83, 79 hybrid; of Lyft's 102, 92 hybrid; of Notion's 76, 49 hybrid and 26 remote.

Why a four percent correction matters

Four percent of remote listings is a small share of the corpus and a large share of one kind of search: the reader who filters for remote and salary. The nine employers above state pay on their listings, so those listings sat high in every remote-and-pay view. Moving them to hybrid changes the remote pay-disclosure rate going forward; the daily series on the pay-transparency page carries a dated basis note from 9 November and is not restated backwards. The weekly digest's section on listings that read remote but describe an office continues; it will now show what the classifier still misses, and readers who find one can reply to the digest.

Method and limits

Basis for the 95: every live listing classified remote on 6 November 2026 (2,344 at snapshot; 18 expired during the two-minute fetch and were excluded; n = 2,326), full description text read from each listing page, parsed with the production office-day parser of that date. Limits: the parser of 6 November read a comma-joined weekday list ("Monday, Tuesday, and Thursday") as one day, so the day-count distribution of that read understates the days for 61 of the 83 listings; the list handling was fixed on 8 November and the per-listing day counts shown on the site are those of the current parser. The quoted sentences were read from the live pages on 10 November; some of the listings counted on 6 November have since expired and others have been posted, so the per-employer live counts of 10 November are not the same set as the 6 November 95. Class counts come from the browsable index's hybrid-mix facet, live basis, read on consecutive days; a one-day comparison includes ordinary turnover, which is why both the remote count and the total are given. The index does not judge whether a schedule is right; it classifies what the listing text states, with the method and its date published. Corrections are appended below, never edited silently.