How a line gets traced, and how we say what we know.
Code Black is archival lineage research for Black American families. Every claim is tied to a record you can open and carries a confidence tier. This page is the method, in the order we actually work it.
1. The record pool
The engine holds 3,167,062 slavery-era record rows across 66 datasets, indexed for meaning (vector search) and for the way enslaved people were actually recorded: by first name under a holder, by holder surname, by place. The largest sets:
| Collection | Rows | What it names |
|---|---|---|
| Enslaved.org people records | 1,223,416 | Named enslaved and free people of colour across the Atlantic world, with the dataset each came from |
| Journal of Slavery and Data Preservation (JSDP) datasets | 771,369 | Peer-reviewed, cited datasets of named individuals |
| Freedmen's Bureau, 1865–1872 (NMAAHC / Smithsonian transcriptions) | 346,632 | Labor contracts, marriages, complaints, rations, hospital and school rolls — the first federal records naming freedpeople |
| Suriname and Curaçao registers | 192,701 | Dutch-colonial enslaved and manumitted persons |
| Liberated Africans (registered + Havana) | 123,945 | People taken off captured slave ships after 1808, often with African names and origins |
| Afro-Louisiana History and Genealogy (Gwendolyn Midlo Hall): Louisiana Slave Database + Free Database | 108,793 | Named people in Louisiana notarial, sacramental and court acts 1719–1820: name, age, sex, origin/ethnonym, holder, parish, the document itself |
| Oceans of Kinfolk (New Orleans slave-ship manifests) | 63,506 | People shipped in the domestic trade, by name, age, sex, shipper and consignee |
| SlaveVoyages: trans-Atlantic and intra-American voyages | 76,374 | Voyages, ports, regions of embarkation — the corridor layer, never a person |
| Accomack County, Virginia register of free Negroes; DLAS petitions; Maryland chattel and slave statistics; Virginia valuations; Franklin & Armfield; New York / New Jersey prices; Information Wanted ads | ~40,000 | County-level registers and petitions naming individuals, holders and kin |
Census indexes (1870, 1880, 1900–1950) are licensed to Ancestry and FamilySearch and are not in the pool. Every result that needs one links you to the exact free FamilySearch search to run. The full coverage table, by place and years, is on the API docs at codeblack-api.dajai.io/docs.
2. The tiers on every claim
No claim leaves the engine without one of these words. They are carried into the report, the API and the GEDCOM export (as QUAY 3 / 2 / 1 / 0).
| Tier | Means | In plain words |
|---|---|---|
RECORD | A primary record names this person and this fact | found |
TRIANGULATED | Two or more independent records agree; none states it outright | found (several records agree) |
TESTIMONY | The family says so; not yet in a record | family said |
COMPILED | Taken from an earlier compiled tree or research file, not re-sourced | reported |
HYPOTHESIS | A lead that fits place, time and name; not proven | maybe |
WITHDRAWN | Was claimed, then disproved; kept so the mistake is visible | no longer held |
Searches that came back empty are kept and delivered too ("looked, not there") — a negative with the record id is worth more than silence.
3. The order of work through the 1870 wall
- Place the person in 1870 and 1880. The first federal censuses that name formerly enslaved people. Race, birthplace, household and the neighbours on the same page all matter.
- Holder candidates from the 1860 and 1850 slave schedules. These list each holder with the age, sex and colour of every person held — no names. The colour mark (B or M) is only on the image, so we open the image. A family present under a holder in 1860 and absent in 1850 was bought or inherited in between: that sale or succession is where the first names are.
- The holder's papers. Successions, inventories, conveyances, notarial acts, sacramental registers, probate. In Louisiana this is where Hall's database and the parish clerks meet.
- Freedmen's Bureau, 1865–1872. Marriages, labor contracts, complaints — often the first time the family's own words appear, sometimes with the former holder named.
- Colonial and antebellum Louisiana by name, origin and holder in Hall's Louisiana Slave Database and Free Database, read in the French and Spanish the clerks used (ethnonyms like Mandinga, Yalunka, Congo).
- The Atlantic crossing as a corridor, not a person. SlaveVoyages gives the voyages that landed a port in the years that fit; an ethnonym in an act narrows the region of origin. We say "region", never "village", unless a record does.
- Everything is logged — hits, misses and the next record to pull — and a human reviews before anything is delivered.
The method worked person by person on a real line is public: the Stewart-Handy case study and the engine's own case file with every record and every miss.
4. What we never do
- Invent a generation to make a line longer. Where the records stop, the report says so.
- Close an open death date or a parent link with a guess.
- Publish living people, children, addresses, estate values or health details. Living persons are withheld from every public surface by construction.
- Give legal advice, or describe any report as evidence for a claim. Code Black is heritage research.
- Bulk-copy a licensed index. We link you to the original.
5. Where the data comes from, and how it is credited
Government and archival records are public domain. Compilations are credited to their makers and linked to the original: Gwendolyn Midlo Hall's Afro-Louisiana databases, the Smithsonian's Freedmen's Bureau transcription project, Enslaved.org and the Journal of Slavery and Data Preservation, SlaveVoyages (whose imputed fields are non-commercial and are not resold), the Digital Library on American Slavery, and the county registers named above. Code Black's own layer is the linking: the match across sources for one family, with the tier and the citation attached.
6. Two numbers you will see, both true
Eight generations documented by record is the showcase line as the engine can cite it today, person by person. Sixty generations is the ancestral reconstruction in the research sessions: the population and its political world at each depth back to the Tichitt tradition, sourced to Al-Bakri, the Tarikh al-Sudan, the Koumbi Saleh and Jenne-Jeno excavations and peer-reviewed ancient DNA. It reconstructs a people, not sixty named individuals, and it says so.