Methodology

The dataset behind this site is three record types, each in its own file:

• Tracked posts — public posts published by @LayoffAI, each paired with dated readings of its public view counter.
• Amplifications — public reposts, quote posts, replies, mentions, citations and interviews by named people or organizations outside the LayoffHedge account itself.
• Media references — external articles, newsletters, podcasts, broadcasts and other public coverage that mentions LayoffHedge or @LayoffAI.

The archive is manually curated by the maintainer from publicly visible sources. It is not an automated crawl and does not claim to contain every public post, repost or article that exists about LayoffHedge — only the ones that have been found and verified.

View counts on this site are readings of the public view counter each X post displays, taken at a specific point in time. Early readings were copied by hand from the post itself, where X rounds large numbers. Later ones come from X's developer API, which returns the same counter unrounded. Neither is a private analytics figure.

Because a public counter keeps changing after it is recorded, every stored view count is paired with the date it was observed — the day the snapshot was taken, not the day the post was published. A post's real, current count may be higher than the value shown here. A repost, quote post or citation recorded on this site is documented amplification, not proof that the person or organization amplifying it endorses, agrees with, or has independently verified any claim inside the original post.

Every record in this dataset carries one of three statuses:

• Verified — the maintainer checked that the record's linked public source (the X post, article or filing) supports the record as written. This means the source was checked by the maintainer. It does not mean verified by X, audited by a third party, or endorsed by LayoffHedge in any way.
• Needs review — the record is incomplete, disputed, or not yet confirmed against its source. It is excluded from every public metric and every public list on this site.
• Archived — kept for the historical record (for example, a superseded or corrected entry) but excluded from current public metrics.

Only records marked verified, and not marked as a development placeholder, feed any number shown on this website. As of September 23, 2026, 100 of the 109 media records in the dataset are verified and visible on this site; 9 are archived, each with its reason recorded, and 0 remain at needs_review, pending manual confirmation, and are counted nowhere.

Nothing on this site is stored as a summary. Every metric below is calculated from the current verified records at build time, using the same functions that render the homepage, the archive and the evidence pages.

• TRACKED POSTS — the count of verified, non-placeholder posts. Today: 32.
• POSTS ABOVE 1M / 5M / 10M — the count of verified posts whose recorded view count is at least that threshold (inclusive, ≥). Today: 17 above 1,000,000; 0 above 5,000,000; 0 above 10,000,000. The archive's filters and the homepage Attention Grid only ever offer a threshold when at least one verified post currently meets it — a threshold with zero matching records is never shown.
• MOST VIEWED TRACKED POST — the single verified post with the highest recorded view count. When two or more posts share the same view count, the tie is broken deterministically and in this order: the post published earliest wins; if the publish dates also match, the post whose record id sorts first, alphabetically, wins. This is the same ordering rule the Viral Archive uses for its row ranking, so "the most viewed tracked post" and archive row 01 are always provably the same record, never a coincidence of two separate implementations agreeing. Today that post has 4,585,209 observed views, recorded on September 21, 2026.
• OBSERVED VIEWS ACROSS TRACKED POSTS — the sum of every verified post's recorded view count. Today: 44,592,802. This total is a sum of per-post counters, not a count of people: the same person may appear in the counter of more than one tracked post, so this number is never a unique-audience figure and is never labeled as one.

Amplification and media counts follow the same rule — only verified, non-placeholder records. An amplifier is counted once as a unique amplifier no matter how many times they amplified the same account; today that resolves to 20 verified amplifications from 20 unique amplifiers, and 100 verified media references across 56 publications. Crossover category counts and Public References counts are the same style of derived count, grouped by category or publication instead of summed.

Some of the coverage in this dataset is a publication's own reporting. Some of it is one outlet republishing another outlet's piece, which wire services, aggregators and partner sites do routinely. Every verified media record says which of the two it is, and a republication names the outlet it credits. Of the 100 verified references today, 81 are the publication's own work and 19 are republications of a piece recorded elsewhere.

That distinction changes what a number means, so two figures are kept apart rather than merged. A count of records — the media figure in the dataset summary and on the evidence page — includes republications, because each one is a real page a real outlet published and a reader can open it. A count of reporting counts originals only, because a republication is the same piece travelling, not a second newsroom reading the data. One outlet's article and eight republications of it are one piece of reporting, and any figure that added them together would overstate the coverage by eight.

Each record also carries the country of the publication's own newsroom — not the country the story is about. Counted across original references only, the verified records come from newsrooms in 7 countries. Where a publication's newsroom country is not settled by a public, citable statement, the field is left blank and that record is counted in no country at all — a blank is not a guess, and it is never quietly filled with the likeliest answer.

Featured references. A reference is marked featured when two things are true of it: the publication produced the piece itself rather than republishing someone else's, and the piece names LayoffHedge or @LayoffAI in its own text — or, for a broadcast, on air — as the source of data or findings it reports, rather than only embedding or linking a post alongside its own reporting. Today 37 of the 100 verified references meet it.

This is curation, and it is declared as curation rather than dressed up as a measurement. It is a rule this project wrote, applied by hand, and it is binary — a reference meets it or it does not. There is no score, no tier and no ranking of publications anywhere in this dataset. All the flag does is position: a featured reference's publication sorts above others that carry the same number of references. There is no star, badge or icon marking it, and a reference that is not featured is not a weaker source or a lesser publication — it means the record does not show the article naming the project as a source in its own words. A reference can only be featured once it has been verified.

A publication that references LayoffHedge is doing one of two things. It is using something LayoffHedge published — the layoff data, the H-1B filings data, an investigation — or it is embedding or quoting an @LayoffAI post and reporting what the post says. Those are different acts, and every verified media record says which one it is. The answer is stored as one of a fixed set of values rather than as free text, so it can be counted rather than read.

A record names a work only when the record's own evidence shows the piece using it: the article names or links a dataset, or one of the surfaces built on it, or it names an investigation. Where the evidence shows an embedded post, a quotation, or LayoffHedge named as a source and nothing further, the record says that no work was named. That is a determination, not a blank — it is the conservative default, the same way a reference that is not featured is not thereby a weaker source.

Counted across the 81 original references — a republication carries the answer of the piece it copies, so counting it too would report one newsroom's use of a dataset as two — 11 cite the layoff data, 12 cite the H-1B filings data, 4 cite an investigation, and 54 name no work beyond the post or the account itself.

Nothing here describes or rates what LayoffHedge makes. The value is an attribute of a reference that already exists, and the list holds only the works these records give evidence of — something LayoffHedge publishes that no verified reference has cited does not appear on it at all, and neither does any judgement about which work matters. The list is closed on purpose: adding to it is a change to the schema and to this page, made when a record shows a work that has none, never a phrase typed into a record.

Every record change follows the same path: identify the right file, add or update the record, attach public evidence for it, normalize the values (ISO dates, raw view counts, never a formatted string), run the data validator, and open a pull request for review before it merges and the site rebuilds. Records added from an automated import (for example, a press sweep) land at needs_review by default and are promoted to verified individually, by hand.

The exact workflow — evidence requirements, id conventions and the full local check sequence — is documented in this project's CONTRIBUTING guide. It lives in the public repository, and a correction or a new record is submitted there; the links are on Evidence & Sources. See About for the project's current status.

• No access to private LayoffHedge analytics of any kind.
• No unique-person or unique-device data anywhere in this dataset — every view number is a public counter snapshot, never a reach or audience figure.
• No claim of exhaustive coverage — the archive is manually curated and may not contain every public post, repost or article that exists about LayoffHedge.
• No endorsement claim — a repost, quote post or citation recorded here documents amplification, not agreement, and political and media records are described factually, never as an endorsement, ranking or validation of any individual or publication.