850 Layoffs a Day: Your Ex-Big-Tech Shortlist Has a 9-Day Half-Life

If you staffed a role off a Boolean-saved LinkedIn project last week, your shortlist is already stale. The 2026 layoff cadence has crossed a threshold where the cut cohort refreshes faster than public profiles update, and static lists no longer track the market they were built against.

Skillsyncer's live tracker shows 365 layoff events and 209,032 workers cut year to date in 2026, an average of 850 job losses per day. That is up from 564/day in 2025, a 51% jump in daily pace. If you built a Boolean-saved shortlist on Monday, roughly 5,950 fresh candidates joined the market before your Friday sync.

Why the "ex-Big-Tech" shortlist decays in nine days

A shortlist decays because the cut cohort refreshes faster than public profiles update. At 850 layoffs per day, the pool grows by about 7,650 people every nine days, roughly the size of a typical Boolean-saved LinkedIn project. Your list has not shrunk; the market around it has moved.

Here is the math in one line: 850/day × 9 days = 7,650 fresh candidates. That is the "half-life" in the title. Not candidates churning out of your list, but new laid-off engineers entering the market who your list never had a chance to see.

The mechanism is worse than the number suggests. Profile updates lag separation by weeks. In Refolk's index, only 383 US engineers surface for "ex-Oracle" against Oracle's ~30,000-role cut, a 1.3% findability ratio. Meta, whose big waves happened earlier in the year, sits at 9.4%. Recency inversely correlates with search visibility, which means the freshest cohort is the least searchable one.

850

The visibility gap: 30,000 cut, 383 searchable

The gap between announced cuts and keyword-findable engineers is 1.3% for Oracle and 9.4% for Meta. If you source on "ex-Oracle" today, you are seeing less than 2% of the real pool, and the 98% you cannot see is the freshest.

Two things fall out of this table:

• Oracle's cut is deep, senior, and invisible. Oracle booked $1.84B in severance and exit costs for fiscal 2026, up from $374M the prior year, roughly a 5x jump. The pool of ex-Oracle senior ICs in Austin, Santa Clara, and Sunnyvale is the largest under-mined cluster of the year.
• "Open to work" is not a strategy. Only 211 US engineers in Refolk's index carry the green banner against 209,032 laid off. That is a 0.1% self-declaration rate. Anyone building a pipeline on the banner is fishing in a puddle and competing with every other recruiter in the same puddle.

Boomerang pollution breaks your Boolean

In the Refolk sample of top "ex-Meta" profiles, 16 of 25 still list Meta as current employer. These are boomerangs, contractors, or profiles that public search mis-tagged. If your string is "ex-Meta" OR "formerly Meta" OR "previously at Meta", you are pulling somewhere between 30% and 60% noise before you even read a headline.

The fix is not a better Boolean. It is not searching on employment history strings at all. Availability signals are downstream of employer changes, tenure gaps, and activity spikes, not headline text.

What a live index does that a saved search cannot

A live index re-scores candidates against your criteria every time you ask, using the current state of GitHub, LinkedIn, and the open web, not a snapshot from when you saved the query. That is the difference between a subscription and a screenshot.

When Oracle cuts another 500 people on a Tuesday, a saved LinkedIn project does not know. It will surface those 500 whenever their headlines update, days or weeks later. This is the exact gap Refolk closes: you describe the person in plain English ("senior backend engineer, ex-Oracle Cloud Infrastructure, US, last title Principal or above") and get a ranked shortlist built against the index as it exists right now, not as it existed when the search was saved.

The public trackers (Skillsyncer, TrueUp, Layoffs.fyi, Crunchbase Tech Layoffs Tracker, plus thelayoff.com for pre-announcement rumour) are the public layer. They tell you that a cut happened. They do not tell you who is now reachable. That translation, from event to person, is what static sourcing tools do badly and a live index does natively.

Your shortlist did not shrink. The market around it grew by 7,650 people while you were in standup.

The five ways static lists lie to you in 2026

Static lists lie in predictable ways once daily layoff volume crosses ~800/day. Here are the five to watch:

1. Recency inversion. The most recent cuts are the least searchable. Oracle's 1.3% findability vs. Meta's 9.4% is the clearest signal. Your list over-represents old cohorts.
2. Boomerang pollution. 16 of 25 top "ex-Meta" hits in Refolk's sample still list Meta as current employer. Any keyword-driven ex-employer search is 30% to 60% noise.
3. Green-banner puddle. 211 open-to-work engineers against 209,032 laid off is 0.1%. The banner selects for the least competitive candidates, not the best.
4. AI-narrative miscoding. The share of cuts blamed on AI rose from ~7% in January to 40% by May, faster than the technology advanced. Many "AI-replaced" ICs are strong engineers whose managers needed cover. Do not down-weight them.
5. Europe blind spot. Sweden shed 1,900 and the Netherlands 1,700 tech workers in early 2026, each more than India's 920. If your saved searches are US-only, you are missing a concentrated senior EU pool freed by continental SaaS cuts.

The signals that actually predict availability

Availability is best predicted by employer-of-record changes in the last 30 days, tenure gaps opening on the current role, and GitHub activity spikes in personal repos. Not by the "Open to Work" banner.

Three signals worth building into any 2026 sourcing workflow:

• Employer transition in the last 30 days. The Refolk index re-checks employer of record on a rolling basis. A candidate whose "current" employer field just flipped is more actionable than a candidate who has carried "Open to Work" for three months.
• GitHub activity spike on personal repos. Laid-off engineers push to side projects. A 3x jump in personal-repo commits in the two weeks after a company's public WARN filing is a strong pre-declaration signal.
• Tenure gap in the current role. A profile whose current role ended 2 to 8 weeks ago and has not yet updated to a new employer is the exact window where reply rates peak. After 90 days, the same profile is competing with 30 recruiters instead of three.

None of these are Boolean-string queries. They are structured questions against a live index. Ask Refolk something like "ex-Uber platform engineers whose GitHub activity tripled in August 2026" and you skip the ghost-hunting entirely.

Where the deepest under-mined pools sit right now

The biggest under-mined pools of Q4 2026 are ex-Oracle senior ICs in Austin and Santa Clara, ex-Uber platform engineers from the July to September 3,300-role cut, and ex-Wix senior full-stack in Israel from Wix's largest-ever 1,000-person cut off a base of 5,280.

By sector, the concentration is unambiguous:

Two named opportunities the trackers under-index:

• Cognition (Windsurf acquirer) laid off 30 employees last week and is offering buyouts to the roughly 200 remaining. These are AI-native engineers, the highest-signal cohort of the year, and almost nobody is sourcing them because the cut is too small for public tracker headlines.
• Uber's 3,300-role cut cites "removing layers, simplifying team structures." Translation: senior ICs and staff-plus managers, not junior generalists. That is the most valuable band of the layoff distribution and the hardest to catch with keyword search because their headlines are stable and prestigious.

What to do differently this quarter

Stop maintaining Boolean-saved projects as your primary pipeline. Treat sourcing as a query you re-run against a live index each time you need a shortlist, with availability inferred from behavior, not banners.

A practical rework of the weekly recruiter loop:

1. Kill the saved LinkedIn project as a source of truth. Keep it as an archive. Do not staff a role from it.
2. Re-query the market at the start of every sprint. With 850 cuts per day, a week-old query is a week-old market snapshot.
3. Replace "ex-[Company]" strings with plain-English descriptions of the person. "Staff engineer who worked on Oracle Cloud Infrastructure networking, cut in the last 60 days" beats any Boolean, and it is exactly the query shape Refolk was built for.
4. Weight behavioral signals over declared signals. GitHub commit spikes, employer-of-record changes, tenure gaps. Not the green banner.
5. Cover Europe. Sweden and the Netherlands each outpaced India in early 2026. If your ATS defaults to US-only, change the default.

Nearly one million tech jobs have been eliminated globally since 2021. 2026 is not the aberration; it is the accelerated version of the same trend. The recruiters who win Q4 stop treating their shortlists as assets and start treating them as queries.

FAQ

How is the "9-day half-life" calculated?

It is a shortlist relevance decay figure, not candidate churn. At 850 layoffs per day, roughly 7,650 fresh candidates enter the market every nine days. That is close to the size of a typical Boolean-saved LinkedIn project, so nine days is the point at which your list is missing a project-sized cohort it never had. The 9.4% ex-Meta findability ratio in Refolk's index implies a similar lag before a fresh cut cohort becomes searchable via public keywords.

Why is "Open to Work" a bad signal in 2026?

Because only 211 US engineers in Refolk's index carry the banner against 209,032 laid off year to date, a 0.1% self-declaration rate. The banner selects for candidates who are actively marketing themselves, which means they are also the most contacted. Behavioral signals like employer-of-record changes in the last 30 days, tenure gaps, and GitHub activity spikes catch a much larger and less competed-for pool.

Should I discount candidates whose roles were "replaced by AI"?

No. The share of 2026 cuts blamed on AI rose from ~7% in January to 40% by May, faster than the underlying technology advanced. That gap suggests AI is functioning partly as a narrative that managers use to justify restructuring. Many "AI-replaced" ICs are strong engineers, and their skills have not degraded because their old manager needed a cover story.

What is the fastest fix for a stale Boolean-saved project?

Rewrite it as a plain-English description of the person you want, and re-run it against a live index instead of a snapshot. Something like "senior backend engineer, ex-Oracle OCI, US, last title Principal or above, separated in the last 60 days" is the shape that works. Refolk was built for exactly this query pattern across GitHub, LinkedIn, and the open web, and it will surface the ~98% of the Oracle cohort that keyword search cannot see.