Amazon's 13/12/12 WARN: Why "Senior+" Filters Miss the Best 37

Amazon's August 28, 2026 Washington WARN filing cuts 121 employees between October 1 and October 27. Thirty-seven of them are software development engineers at levels I, II, and III, split 13/12/12. That flat seniority profile breaks the "Senior+" LinkedIn filter most sourcers reach for first, and it does it inside a 27-day separation window that starts the same morning the October HN "Who is Hiring" thread opens.

What the filing actually says

Amazon is separating 121 Washington employees between October 1 and October 27, 2026, across 12 named facilities in Bellevue (53), Seattle (35), and Sumner (32), plus one remote worker. The WARN letter is dated August 28, 2026; Washington's Employment Security Department received it August 31; GeekWire broke it the same day.

The structure that matters for sourcing:

• 52 of the 121 roles (43%) are software development engineers or applied scientists.
• 37 of those are SDE I, II, and III, split 13/12/12.
• First separation lands October 1. By October 2, 32 people are out. By October 20, 66. Last separation is October 27.
• Affected workers got 90 days' notice and can apply for internal transfers before the cuts become final.
• This is Amazon's fourth Washington WARN of the year, bringing announced WA reductions to more than 4,600 since last fall.

The piece most sourcers will miss: the WARN specifies levels. That is rare. Most WARN filings list headcount against generic titles. This one hands you a level distribution, which means you can back out exactly where to look.

Why the "Senior+" filter is wrong this time

The default LinkedIn filter for ex-Amazon engineers is Senior and above, because mid-level is where the stock sits. For this filing, that default throws away roughly two-thirds of the pool. The 13 SDE-Is are not a rounding error; they are the densest slice of a tiny national segment.

Here is the shape of the pool, with the WARN headcount against the national stock in Refolk's index of professional profiles.

Now the two derived numbers that should change how you work the week of October 1:

1.89%

The SDE-II line is the mirror image: 12 out of 7,041 is 0.17% of national stock. Interesting, but not scarce. The SDE-I slice is scarce. A sourcer who filters it out is skipping the rarest segment on the list, not the least valuable one.

The mechanism behind the flat cut

Historically, Amazon's engineering layoffs skewed mid-level. Stocktwits summarized the prior round plainly: "software engineers, particularly those in mid-level roles, known as SDE II, were among the hardest hit." This filing is different. Amazon is pouring $220 billion into AI infrastructure this year, a narrative that treats junior engineers as the layer most substitutable by Copilot-class tooling. So the cut flattens across levels instead of concentrating in the middle. The pattern shifted. The default sourcing filter has not.

The 90-day internal transfer wrinkle

The 90-day notice is not a gift to the laid-off engineer; it is a filter on who is actually available to you. The best SDE-IIIs will transfer internally before October 27. The SDE-Is mostly cannot, because they do not have the internal connections or org capital to land a transfer in 60 days.

Translate that into a sourcing hierarchy:

1. SDE-I (13 people). Highest actual-availability rate. Lowest internal-transfer success rate. The segment every "Senior+" filter deletes.
2. SDE-II (12 people). Mixed. The strong ones with an internal network move laterally into AWS or Ads. The ones outside hot orgs hit the market.
3. SDE-III (12 people). Lowest actual-availability rate. Many will land an internal transfer, often into AI infrastructure teams spending some of the $220B.

The counterintuitive consequence: a disciplined sourcer should spend more time on the SDE-I bucket than on the SDE-III bucket during this window, which is the inverse of the usual effort split.

The geography compresses the whole cohort into a week of coffees

Thirty-seven engineers across 12 buildings in Seattle, Bellevue, and Sumner is roughly 3 per building. Combined with Refolk's index showing Seattle as the top region for SDE-II profiles (40% of the top-25 sample), a single in-person loop can realistically cover the entire cohort in a week. That is almost never true of a WARN. Most WARN pools are spread across the country, and in-person loses to InMail at scale.

Here it does not. A Bellevue-based recruiter can:

• Hit three Downtown Bellevue coffee shops on Oct 2, catching the 32 people out by then.
• Add a South Lake Union loop on Oct 6 for the Seattle contingent.
• Run a Sumner site visit on Oct 15.
• Close by Oct 20, which is when the Bay Area wave lands and the LinkedIn graph gets noisy.

The Bay Area wave is a 19-day countdown

Amazon filed a second WARN in California the same August day: 148 roles, 78 in San Francisco and 70 in Sunnyvale, all separating October 20, 2026. Washington's separations begin October 1. That is a 19-day head start for recruiters working the Puget Sound slice before the Bay Area ex-Amazon graph gets flooded by SF and Peninsula sourcers chasing the same alma mater signal.

The two-state total from a single day of paperwork is 269 people. The sourcers who win the SDE-I piece of that will be the ones who stop filtering by level and start filtering by separation date.

The HN "Who is Hiring" collision

October 1 is also the day the monthly Hacker News "Who is Hiring" thread opens. The supply side (37 named, dated, Seattle-dense Amazon engineers) and the demand side (startup posts with founder emails, no ATS) hit the same morning. The right play is to pair them explicitly: pull the hiring posts that want Python, distributed systems, or AWS experience, and warm-intro them to the SDE-I/II slice of the WARN cohort the same week.

A playbook for October 1 to October 27

Treat the window as four phases. Each phase has a different highest-value action.

Oct 1 to Oct 6: lock the SDE-I bucket

The 13 SDE-Is have the lowest internal-transfer rate and the smallest national pool. First-touch wins:

• Pull the Amazon Seattle/Bellevue SDE-I subset from your tool of choice.
• Cross-reference against the WARN's named buildings (available in the filing).
• First message by Oct 3, before the Oct 6 reflex spike in outreach.

Oct 7 to Oct 19: work the SDE-II middle

• Filter for SDE-IIs whose team names on LinkedIn match orgs Amazon has publicly deprioritized.
• Skip SDE-IIs on AWS core teams; they are almost certainly getting the internal transfer.
• Prioritize anyone with a GitHub commit history outside Amazon's monorepo; they are more likely to interview quickly.

Oct 20 to Oct 24: brace for the Bay Area wave

• October 20 is when 66 WA separations and 148 CA separations land. LinkedIn becomes a shouting match.
• Your advantage is a prepared SDE-III shortlist with named roles and a scheduled interview loop, not another InMail.
• Any SDE-III still available on Oct 20 is one who already tried and failed to secure an internal transfer, which is useful signal.

Oct 25 to Oct 27: the stragglers

• Final separations hit Oct 27.
• The last-week pool skews toward people who had the hardest time landing anywhere, which is a mix of genuine misses and the people the market mispriced.

Any SDE-III still on the market October 20 already lost the internal-transfer race. That is signal, not noise.

What to subscribe to so you do not need the press

The press is 24 to 48 hours late on WARN filings. Three sources beat it:

• Washington State Employment Security Department WARN system. Publishes filings directly; you can poll it or scrape it.
• GeekWire. Fastest same-day coverage for Puget Sound tech WARNs. Breaks the context (building names, prior rounds) the raw WARN does not.
• Amazon Rapid Response / WorkSource. State job-placement teams contact impacted employees to help with the transition. If you understand their outreach cadence, you know when candidates have been primed for a conversation.

The broader WA context matters too. The state's Employment Security Department received at least 21 WARN filings from information technology employers in 2025 affecting up to 6,700 workers, concentrated in King County and dominated by Amazon and Microsoft. The same quarter has seen large WA layoffs from Microsoft, Zillow, Starbucks, Google, Oracle, and Meta. The ex-Amazon pool is competing for the same seats as everyone else's cuts.

The one-line rule to take away

When a WARN specifies levels, treat the level distribution as structured data, not news. Filter by separation date and employer, not by seniority. The SDE-I slice of Amazon's October WARN is the clearest demonstration you will get this quarter that the default "Senior+" filter was written for a labor market that no longer exists.

FAQ

How do I confirm which Amazon buildings are in the WARN?

The WARN letter dated August 28, 2026 names 12 facilities across Bellevue (53 employees), Seattle (35), and Sumner (32), plus one remote worker. The full address-level list sits in the filing on the Washington ESD site. Pull it directly; press summaries round the building list.

Why does the WARN include SDE levels at all? Most do not.

Washington's WARN form asks for job titles, and Amazon chose to use its internal level nomenclature (SDE I, II, III) rather than a generic "software engineer" bucket. That is a company-by-company choice. When levels appear, treat them as a gift and build your pool from them directly.

Is 37 engineers really worth a dedicated playbook?

For sourcing Amazon SDE-Is, yes. The 13 SDE-Is in this filing are roughly 1.89% of the entire US SDE-I-titled talent pool in Refolk's index, in one 27-day window, in one metro. The 12 SDE-IIs are rounding error against national stock, but they are named, dated, and in-market, which is what makes any WARN pool valuable.

How does the parallel California filing change the Seattle math?

The same August filing wave includes 148 California roles separating October 20, 2026 (78 San Francisco, 70 Sunnyvale). Bay Area recruiters will saturate the ex-Amazon LinkedIn graph starting October 20. Washington recruiters who move on the October 1 separations have a 19-day head start before the signal-to-noise ratio collapses.