Florida — Operation Statewide Shield (Aug 17–21, 2026)
FDLE and the Florida Attorney General's office announced the recovery of 163 missing children (ages 2 months to 17) in a five-day statewide operation — 46 in Tampa Bay, 41 in Miami, 32 in Jacksonville, 3 in Tallahassee, with sub-operations in Pensacola, Orlando and Fort Myers; many were trafficking victims. No names were released, and recovered children are removed from the active poster list — so they do not appear below. Florida children reported missing before Aug 22 who are still listed are flagged still listed after Shield: they were not among the 163.
Sources: WUSF, Aug 25 2026 · WFLA · Florida AG release · earlier: U.S. Marshals Operation Dragon Eye (60 children).
Filter
Where they went missing — circle size = children per state; click a state to filter
Children
| Name | Age · sex | Race | Missing since | From | Type | Source |
|---|
| # | Name | Category | Status | Research |
|---|
Case locations
Hanging deaths — map, timeline & sequence — every tracked case found hanging or with a ligature, in date order
| # | Date | Time found | Name | Age · sex | City, state | Setting | Weather at the site (day of) | Where / how found | Official ruling | Family position | Days since prev. | Miles from prev. | Sources |
|---|
Manner of death
Official ruling
Race / ethnicity
By state
By year — most recent first, back to 1979
Documented concerns
Missing children & missing persons — the tracker's own missing-person cases, missing → found
| Name | Reported missing | Missing from | Found | Found at | Days | Outcome |
|---|
Missing children registry — every active poster for a Black, Indigenous, Latino or Asian child, live from NCMEC and the BIA
① Collection trigger — does this death qualify?
② Deliberate search builder
③ Lead queue — articles found
| Found | Person / headline | Location | Score | Status |
|---|
④ Sweep algorithm — signature search
Derived from the cases in this tracker. Six signatures describe how these deaths surface in coverage — the words a reporter uses before anyone calls it a pattern. The sweep runs every signature phrase against a demographic term and the time window set in ②.
- Cadence. Weekly: set ②'s window to "Past week", run the sweep, open each link (Google News first; GDELT catches wire and small-outlet pickups).
- Screen. Read each hit against the gate and indicators in ①. Score ≥ 2 → log as a lead; score ≥ 4 → promote to a case.
- Dedupe. Match by normalized name + state. The Names and Cases tabs already flag anything present.
- Enrich. For each lead run "Name" city plus autopsy · ruling · arrest · footage · attorney · NAACP. The Names tab's per-name links do this in one click.
- Watch-back. Re-run each open case's name at 7, 30, and 90 days: rulings change, charges land, independent autopsies and FBI/DOJ reviews appear late.
- Cross-check. JULIAN's MDL database / A Crimson Record, Mapping Police Violence, and Say Every Name.
- Prioritize sources. Black press (Capital B, Mississippi Today, Word In Black, The Miami Times, Kansas City Defender, BIN, The Root, Emerald Book) and local TV (Gray, Nexstar, Tegna, FOX O&Os) break these stories days before national outlets.
⑤ A family member is missing and police aren't acting
- Insist on the report. Police must take a missing-child report immediately (federal law — no waiting period) and enter the child in NCIC within two hours. Ask for the report number and the NCIC entry confirmation.
- Call NCMEC: 1-800-843-5678 (1-800-THE-LOST). They open a case, produce the poster, and push the agency. missingkids.org/gethelpnow
- Black and Missing Foundation — amplification and agency pressure for missing Black people: blackandmissinginc.com.
- Enter NamUs (adults especially): namus.gov — the national missing-persons database; families can create a case that the agency must review.
- Indigenous relatives: BIA Missing & Murdered Unit tip line 1-833-560-2065 · bia.gov/service/mmu; FBI field office.
- Escalate in writing to the chief/sheriff, the district attorney, and the state missing-persons clearinghouse; copy a reporter from the Black press. Keep every message — it becomes the record of delay.
- Preserve evidence yourself: phone location sharing, social accounts, doorbell/business cameras (Cameras tab), and the names of everyone last with them.
Systemic tracking & key disparities
The phenomenon of misclassified or uninvestigated deaths in Black, Indigenous, Latino and Asian communities spans several analytical sectors at once: what the death certificate says (vital statistics), what the coroner or medical examiner actually did (autopsy, toxicology, scene work), and how police counted, classified, and closed the file. No single database holds all three — so disparities are found by comparing systems and by coding each case so it can be lined up against national data.
Two signals matter most. A death parked in Y10–Y34 ("undetermined intent") or R99 ("cause unknown") means the classifying office did not or could not choose between accident, suicide, and homicide. A rate of such classifications that is higher for one racial group than another, in the same jurisdiction and years, is a measurable disparity in investigative effort.
Live tally from your cases
ICD-10 classification
Investigation-quality indicators
Comparative breakdown of raw data systems
| Database | Data source | Access | Best used for |
|---|---|---|---|
| CDC WONDER | Death certificates only — 1979–1998 (Compressed Mortality File, ICD-9 codes) and 1999–present (Underlying Cause of Death, ICD-10); Multiple Cause of Death 1999–last week, incl. provisional. Nothing before 1979 is searchable by race and cause in WONDER. | Public aggregated query builder |
Tracking macro-level spikes in Y10–Y34 "undetermined" and R99 "unknown cause" deaths by race/ethnicity, state, county, year, age, and sex. |
| NVDRS RAD National Violent Death Reporting System — Restricted Access Database |
Death certificates + coroner / medical-examiner reports + law-enforcement reports + toxicology, with circumstance narratives | Restricted application; academic / agency |
Analyzing the absence of toxicological testing, scene details, or circumstance data in deaths of Black, Indigenous, Latino and Asian decedents versus white decedents in the same jurisdiction. |
| NACJD / ICPSR National Archive of Criminal Justice Data, University of Michigan |
FBI Supplementary Homicide Reports (SHR), National Incident-Based Reporting System (NIBRS), Uniform Crime Reports; police incident-level records | Public & restricted public-use + restricted-use files |
Tracking how police classify, clear, or prematurely close files — and finding deaths that appear as homicides on certificates but never in police homicide counts (or the reverse) — before forensic results were returned. |
| EJI — Lynching in America | Historical documentation of racial terror lynchings; interactive map; county data supplement | Public | Historical baseline: the era when deaths of Black people were uninvestigated by design. Lets modern cases be placed against the documented county-level geography of racial terror. |
| data.gov | Federal open-data catalog; indexes state/county medical-examiner, coroner, and police datasets | Public | Locating local "manner of death by race" tables, ME/coroner case logs, and police open-data portals for a specific jurisdiction. |
| Mapping Police Violence | Every killing by police since 2013, compiled from media reports, obituaries, public records, and other databases; race, agency, circumstances, and whether charges followed | Public downloadable dataset |
Cross-checking any death with police contact before it — and comparing how often charges follow when the person killed is Black versus white. |
| FBI Crime Data Explorer Uniform Crime Reporting program |
NIBRS incident data, Expanded Homicide Data (SHR), hate-crime and National Use-of-Force collections; bulk CSV downloads and an API | Public | Agency-level homicide counts and clearance rates by victim race — finding agencies whose reported homicides don't match certificate counts from CDC WONDER. |
| BJS LEARCAT Law Enforcement Agency Reported Crime Analysis Tool |
NIBRS incident, victim, and arrest data from reporting agencies (2024 incident counts linked) | Public interactive query tool |
Whether arrests follow homicides of Black victims at the same rate as others, by agency, year, offense, and victim race / sex / age. |
| VPC — Black Homicide Victimization Violence Policy Center, 2024 data |
CDC WISQARS mortality data (national and state rates) plus FBI SHR (victim–offender relationship, circumstances) | Public report | The national and state baseline for Black homicide — the rate against which every "not a homicide" ruling in this tracker should be read. Figures below. |
| Say Every Name Say Their Names list |
Memorial roster of Black people killed in the U.S. by law enforcement and civilians — names only | Public loaded in the Names tab |
A starting roster to research one name at a time: add as a case, establish date/place/ruling/charges, and cross-check against Mapping Police Violence. |
| JULIAN — MDL database "A Crimson Record" (Feb 2026) |
Documented suspected modern-day lynchings (MDLs): 70+ cases across MS, AL, FL, GA, LA, TN, and TX since 2000 — Mississippi 20 — a 21st-century follow-up to Ida B. Wells's A Red Record | Public report + growing database |
The closest existing analogue to this tracker: match every hanging or "ruled suicide" case here against JULIAN's list, and adopt its criteria when writing up a pattern. |
ICD-10 codes (deaths from 1999 on) that capture uninvestigated or misclassified deaths
| Block | Meaning | Sub-codes worth pulling |
|---|
ICD-9 codes (deaths 1979–1998) — the same categories for the earlier years
| Block | Meaning | Sub-codes worth pulling |
|---|
How to run the comparison
- CDC WONDER → Underlying Cause of Death (1999–present). Group results by Single Race 6 (and Hispanic Origin), State, and Year. Under ICD-10 codes select Y10–Y34. Export. Repeat for X60–X84, V01–X59, and X85–Y09 so you have every manner for the same group/place/years. For 1979–1998 use Compressed Mortality with ICD-9 E980–E989, E950–E959, E800–E949, and E960–E969; race there is the older 4-category coding.
- Compute the share. For each racial group: undetermined-intent deaths ÷ all injury deaths (V01–Y36). Enter two groups in the calculator below to get the disparity ratio.
- NACJD → Supplementary Homicide Reports for the same state and years, filtered by victim race. Compare the police homicide count against WONDER's X85–Y09 certificate count. A gap in either direction is a classification question.
- Local records. Search data.gov for the county ME/coroner. If no "manner by race" table is published, file a public-records request for it — and for the office's toxicology-testing rate by manner.
- Code every case in this tracker with its ICD-10 block, then use the tally above to see how your documented cases sit against the population-level numbers.
Disparity ratio calculator
Group A
Group B
Historical baseline
The Equal Justice Initiative's Lynching in America documents 4,075 racial terror lynchings of African Americans across 12 Southern states between 1877 and 1950 — at least 800 more than previously reported — with follow-up research covering eight additional states (IL, IN, KS, MD, MO, OH, OK, WV). Its interactive map and county data supplement give a documented, county-level geography of deaths that were never investigated. Add the EJI county count as a tag or note on a modern case when the location overlaps.
National baseline — Black homicide victimization, 2024
85.3% of Black victims were male; 84.3% were killed with firearms. Highest state rates: Missouri 48.3, Mississippi 41.0, Illinois and Wisconsin 38.6 per 100,000. Source: Violence Policy Center, from CDC WISQARS and FBI SHR. These count deaths that were classified homicide — every misclassified death is missing from them, which is why the undetermined-intent share matters.
Computed from medical-examiner open data — the disparity ratio, run for real
Denominator is every ME case (natural deaths included), so the shares are small; the ratios are the point. Re-run any time: the API links return current counts.
Medical examiner & coroner public portals
Missing children without a voice — what the numbers show
Sources: WTVR (May 24, 2026) citing NCMEC; Black and Missing Foundation; FBI MMIP. Read carefully: these are shares of reported cases, not proof of a rising count — the Black and Missing Foundation itself said it could not verify the viral "uptick" figures, while confirming the disparity is real. To test an uptick, pull NCMEC's counts for the same months in prior years and compare. In this tracker, filter Cases by the tag missing child or the status Missing (not found).
Modern-day lynching — JULIAN's definition
JULIAN, the first organization to document modern-day lynchings in the 21st century, defines an MDL as a discriminatory killing carried out by more than one person for an alleged offense, with or without legal trial or due process — intended, unlike an ordinary hate crime, to spread terror and uphold a system of supremacy, and often involving an element of spectacle. Its February 2026 report A Crimson Record documents more than 70 suspected cases across seven Southern states since 2000, with Mississippi leading at 20.
The report's central observation is this tracker's premise: lynchings by hanging tend to be ruled suicides at first, and investigations close before they begin. Use the hanging — tree tag and the "Ruled suicide/accident over family objection" concern to pull the matching subset from your cases.
Sources for everything on this tab
Nothing on this tab is estimated by the site. Where a source could not be fetched automatically (Mapping Police Violence, FBI CDE, LEARCAT, several ME portals), the row describes the source from its public documentation and links it; no numbers were taken from those.
Mission
To help locate, and to help prevent, the disappearance and death of more children and people without a voice — Black, Indigenous, Latino and Asian lives. To support and inform those who cannot get the answers they need. To bring back together a community that has been broken by so much heartbreak.
This site is a complete tool as it stands today — version — and it will be updated regularly until the full vision is built (see the roadmap below).
Purpose
A structured record of deaths of people and children without a voice — Black, Indigenous, Latino and Asian lives — that were dismissed, prematurely closed, or under-investigated — deaths ruled suicide, accident, or "undetermined" over family objection, or affected by disputed autopsy findings — regardless of the manner of death (drowning, hanging, fall, and so on).
By logging dates, places, manners, demographics, and investigation quality in one place, the tool lets patterns — including clusters of deaths close together in time or geography — become visible, so families, journalists, attorneys, and oversight bodies can examine them.
Ground rules built into the data model
- Separate the official ruling from the dispute. Every case records what authorities concluded and, separately, what the family/community allege.
- Cite everything. Each case has a Sources list; unsourced claims are marked unverified.
- A cluster is a question, not an answer. The tool surfaces deaths that fall close together but never asserts why. Coincidence, shared conditions, reporting effects, and coordination are all possibilities to be tested, not assumed.
- Describe, don't accuse. Log disputes as allegations unless a source establishes otherwise.
- Code to the national standard. Each case carries its ICD-10 block (ICD-9 for 1979–1998) so it can be compared against CDC WONDER, NVDRS, and NACJD data — see the Disparities tab.
- Missing is a status, not an afterthought. A person reported missing is a record from day one — with where and when they were reported missing kept apart from where they were found — and the outcome (still missing, found alive, found deceased, recovered from trafficking) is tracked explicitly.
The missing-children registry
The Patterns tab carries every active missing-child poster for a Black, Indigenous, Latino or Asian child in the United States — pulled from NCMEC's public poster search by race and state, with each child's detail record (sex, case type, circumstances), cross-checked against Global Missing Kids, and joined by the Bureau of Indian Affairs Missing & Murdered Unit's published cases. Race is exactly as the source records it; Black, Hispanic/Mexican, and American Indian children are highlighted. The registry is a snapshot with a pull date; children who are found are removed from the sources and will drop out on the next pull. Any child can be added to the tracker with one click, after which they cluster and map with everything else.
Florida's Operation Statewide Shield (Aug 17–21, 2026) recovered 163 children without releasing names; Florida children reported missing before it who are still listed are flagged, because they were not among the 163.
Medical-examiner data and the computed ratio
Where an ME office publishes every case with race and manner (Cook County, San Diego County), the site computes the share of deaths classified "undetermined" by race and the ratio to white decedents directly from the office's API — the same measure the Disparities calculator asks you to run on CDC WONDER. Nineteen ME and coroner portals are linked; each case's records section shows the portal for its state or county.
When a family reports someone missing and police don't act
Intake ⑤ logs the person as a missing-person case immediately — not after a death — records what the family was told, and lists the escalation path (NCMEC, Black and Missing Foundation, NamUs, BIA MMU for Indigenous relatives, written escalation, evidence preservation). The delay itself becomes part of the record, which is what the disparity data says to look for.
Camera / evidence catalog
For each case you can log nearby cameras — Flock Safety ALPR readers, DOT/traffic cameras, business CCTV, residential/doorbell cameras, and public cameras — with location, owner/contact, footage-retention window, and request status. The Cameras tab lists where to find them, state by state.
Data & privacy
Everything is stored locally in your browser. Nothing is uploaded. Use Export to back up and share; Import to merge or replace. Seeds reloads the built-in starter set (with a warning).
Official case searches — what is actually public
Every case card links the court-record search for its state (the Court & records links). What those searches return depends on the state. Most have a free public docket (Maryland Case Search, NC eCourts, SC Public Index, Alacourt, WCCA in Wisconsin, MiCOURT, Kentucky CourtNet). Mississippi is different: circuit and county dockets and filed documents are in the Mississippi Electronic Courts system, which the public reaches through a PAMEC account ($10 a year) or free at a courthouse terminal; the Supreme Court and Court of Appeals dockets are free online. Two limits apply everywhere. First, a death with no charges produces no court case: a grand jury's no-true-bill, a medical examiner's ruling, or a police closure leaves nothing on a docket, so the records to request are the ME report, the incident report, and any grand jury report the DA chooses to publish. Second, autopsy reports are confidential in many states (Mississippi, Georgia, and Florida among them) except to next of kin and their counsel, which is why a family's independent autopsy and their attorney's releases are often the only primary documents in public view. The Nolan Wells card shows the pattern: the DA published a redacted grand jury report, the state ME's findings are known only through that report, and no case number exists.
Hanging deaths — how the map and sequence analysis work
The Patterns tab's Hanging deaths section pulls every case whose manner is hanging or whose description mentions a noose or ligature, orders them by date, and numbers them on the map. Each row shows the date, the time the person was found (as published — a 911-call time, a "pronounced" time, or "early morning" when that is all the record says), and the days and straight-line miles from the previous case. The dashed line on the map joins consecutive cases in time only.
The sequence analysis tests one specific question — could a single group have been present at consecutive scenes? — using a generous 600 road-miles-per-day limit. Pairs that fail that test are evidence against one traveling group for those two cases; pairs that pass are not evidence for it, only "not ruled out." The analysis also counts cases where police say video shows the person alone, and cases where a suicide finding was announced before the autopsy. It says so on the page: timing and geography cannot prove coordination; ligature, toxicology, injury, phone-location and license-plate evidence can. Use the Cameras tab and the records-request checklist on each case to go after those.
Weather. Each hanging row carries the weather at that site on the day the person was found and the day before — sky condition, low and high, rainfall, peak wind, sunrise, and the hour-of-discovery reading when a time is known. The source is the ERA5 reanalysis served by Open-Meteo (archive-api.open-meteo.com), a gridded model of observed weather at roughly 9 km resolution, not a station report; for a case in the last few days the site uses Open-Meteo's analysis feed instead. Weather is embedded for the seeded cases and fetched live for new ones when the browser is online. It is route context — road conditions, daylight, how busy an outdoor scene would have been — and the page says so; it does not establish motive.
Activity log & version control — who changed what, when
| When | Action | Case | Detail | Device | Rev |
|---|
Operations settings
Every add, edit, delete, import, and promotion posts a JSON event to the webhook (action, name, detail, device, region, rev, version, text) the moment it happens; "site opened" stays in the local log. The region option asks your browser for permission once, keeps only the nearest state, and applies only to your own entries — visitors are never located. There is no server behind this file, so it cannot see who opens it; the hosted copy's sync is how other people's activity reaches you.
Hosted copy — access logging & notifications — for the S3 / CloudFront site
The settings above are per browser, which is right for a file you hand to people but not for a public site. The hosted copy has three layers, all disclosed in the footer:
- CloudFront standard logs — the authoritative access record, kept by AWS, no code involved: every request with time, IP address, country, URL, browser string, referrer and status. Turn on in the console: CloudFront → the distribution → General → Settings → Edit → Standard logging: On → choose a private S3 bucket for the logs → Save. Logs arrive as gzipped files within about an hour. For a live view, CloudFront → Reports & analytics shows viewers, popular pages, and countries without any setup.
- Access beacon — set SITE_CONFIG.accessBeacon near the top of the page's script to a webhook URL before uploading. Every page open then posts one "visit" event: time, page, referring site, browser type, mobile or desktop, language, time zone, screen size, an anonymous per-browser id and the site version. No names, accounts or precise location. Send it to Slack for a running feed, or to a Google Apps Script or Zapier "add row" for a spreadsheet.
- Site-wide change notifications — set SITE_CONFIG.changeWebhook the same way. Any visitor's add, edit, delete or import then notifies you even though they never opened these settings; a visitor who sets their own webhook overrides it for their own browser.
Both constants are blank in the file you download here, so nothing is sent from a local or webloc copy. Fill them in the copy you upload. The webhook URLs are visible in the page source to anyone who looks — use endpoints that only accept posts (a Slack incoming webhook or an Apps Script web app), never a key with read access.
Version history & roadmap
Roadmap — the full vision
- Server-backed version control and audit trail with named accounts, so every contribution is attributed and reversible.
- Notifications to subscribed families, organizers, and reporters when a case they follow changes or a new one is added nearby.
- Scheduled re-pulls of NCMEC, BIA, and medical-examiner feeds so the registry and the computed disparities stay current without a manual build.
- Hosting on a content-delivery network with a stable address, offline-first on phones, and privacy-preserving usage statistics (counts only, never identities).
- Public-records request templates and a per-case deadline tracker for footage retention windows.
Contact & copyright
Questions, corrections, or a case to add: w7hgnxs6cn@privaterelay.appleid.com
© 2026 Unresolved. All rights reserved. Case summaries are compiled from the cited public sources; official rulings and family positions are recorded separately and every entry is marked verified or unverified. Nothing here asserts guilt of any person.
Ask the tracker's data — answered right here from this site's records; nothing leaves the page
How to write a prompt that gets real answers — guide and ready-made templates
A good prompt has five parts. The builder below adds parts 3–5 for you; you supply 1 and 2.
- Who or what — a full name, a place, a date range, or a pattern. Specific beats broad: "Tasia Fortune, Jackson MS, found Aug 3 2026" outperforms "hanging deaths". Attach a tracker case and its details go in automatically.
- What you want back — one clear ask per prompt: the ruling and autopsy status, whether anyone was charged, the last-48-hours timeline, which cameras were nearby, or similar deaths in the same period. Tick "Also ask for" chips to add more.
- Where to look — pick the sources in step 2. Local TV and Black press break these stories first; court portals show charges; camera maps show what could have recorded the scene.
- The site's own data — turn on "Include tracker data" and the assistant gets the matching records from this tracker (dates, rulings, family positions, sources) and is told to search those first, then the web.
- Rules — always attached: a URL and date for every fact, "not found" instead of guesses, official rulings kept separate from allegations, nobody accused without a charge on the record.
Templates — click one to load it, then edit the [brackets].
Tips. Ask one question at a time. Give dates as full dates. Name the county, not just the city. If the first answer is thin, send the follow-up "List every source you checked and what each said". For social media, search the person's name plus the city and the word "found" in the days after the date.
1 · Your question or prompt
2 · Where should it look?
3 · Send it — each button opens directly in that site or app, with everything pre-filled
Surveillance & ALPR maps
Government & registered cameras — DC / Maryland / Virginia
Public webcams & aggregators
Cameras by state — traffic, public webcams, surveillance maps
| State | Traffic cameras | Public webcams | Surveillance & ALPR maps |
|---|