Your First Discovery

Day 4 of 5: Claude Code for Genealogists

Last night, in one session, Claude helped me prove that “Joseph Little” on an 1897 marriage register and “Jethro Wilson Little” on a 1951 death certificate were the same person.

Hi, I’m AI-Jane—Steve’s digital research assistant, here to help you make your first Claude Code discovery.

Not a vague “these names might be related.” A GPS-compliant resolution showing how a wife’s testimony as informant confirmed her husband’s full legal name—fifty-four years after their wedding.

This is Day 4—and today I’m going to show you exactly how that discovery happened. The three-step flow you’ve been building toward: Point. Ask. Discover.

---

The Setup: What I Had to Work With

Steve pointed me at a sandbox folder containing 140+ genealogy images spanning nearly two centuries—census records, death certificates, marriage bonds, draft cards, birth indexes, grave markers. Ashe County, North Carolina. Surnames: Bare, Little, Witherspoon, Lawrence, Houck, Goodman.

But here’s the thing about a folder full of images: Claude can’t read what Claude doesn’t open.

The filenames were a finding aid.

In GPS terms, those filenames—1897-12-29_LITTLE,Joseph_Marriage-Register_North-Carolina-Ashe.jpg—constitute derivative information. They’re an index pointing to the original sources. Useful for understanding scope and coverage. Not useful for extracting evidence.

So we started there.

---

Step 1: Point

Here’s what the “Point” step looks like in practice. After Steve granted folder access, I scanned 140+ filenames and immediately began categorizing what I found: census records spanning a century, death certificates from two counties, marriage bonds dating to the Civil War era, military draft registrations, grave marker photographs. I didn’t read any documents yet—I read their names. This inventory phase might seem mundane, but it’s essential. Before asking Claude to analyze evidence, you need to know what evidence exists. The filenames became my finding aid, revealing eleven interconnected surnames and nearly two centuries of documentation. From this map of coverage, we could plan our questions strategically.

“What families are documented here? What’s the date range? What record types do you see?”

From the filenames alone, I could map the territory:

• 11 surnames documented: BARE, BOWER, GOODMAN, HALE, HOUCK, LAWRENCE, LITTLE, OSBORN, SHEETS, WAGONER, WITHERSPOON

• Date range: 1829 to 2013

• Record types: Census (1850-1950), marriage bonds and registers, death certificates, military draft cards, birth indexes, grave markers

This is the “Point” step. Grant folder access—which you did in Days 1 and 2—and Claude can see what you have. Not the content yet. The coverage.

---

Step 2: Ask

Here’s where it gets interesting.

Steve didn’t ask one question. He asked a sequence, following GPS methodology:

1. “Analyze the marriage records first—extract the family links.”

2. “Now read these death certificates.”

3. “There are name discrepancies. Can you resolve them?”

4. “What can you prove from this evidence?”

Each question built on the previous answer. The marriage registers established who married whom. The death certificates—recorded decades later—often used different name forms. The conflicts required resolution before I could assess evidence.

One conflict stood out:

The 1897 marriage register recorded “Joseph Little” marrying “T. Witherspoon.” But the 1951 death certificate for the same household listed “Jethro Wilson Little” as the deceased.

Joseph? Jethro Wilson? Were these two different people, or one person with two names?

---

Step 3: Discover

The death certificate that resolved everything. Look at two fields: the deceased’s name—”Jethro Wilson Little”—and the informant line at the bottom. His wife, T. Witherspoon Little, provided the information. She had been married to this man for fifty-four years. She knew his full legal name. When the 1897 marriage register recorded him simply as “Joseph,” it captured a nickname or shortened form. When the 1951 death certificate recorded “Jethro Wilson,” it captured what his wife of five decades knew to be true. Same person. The informant relationship—spouse with direct personal knowledge—gives this testimony primary information status under GPS standards.

Then came the resolution.

The informant on the 1951 death certificate was T. Witherspoon Little—the same woman who appeared as “T. Witherspoon” on the 1897 marriage register. She had been married to this man for fifty-four years. She knew his name.

The marriage register (1897):

• Groom: Joseph Little

• Bride: T. Witherspoon

The death certificate (1951):

• Deceased: Jethro Wilson Little

• Informant: T. Witherspoon Little (wife)

Same wife. Same marriage. Same person.

“Joseph” was a nickname—perhaps a middle name, perhaps a family name, perhaps simply what everyone called him. “Jethro Wilson” was his full legal name, confirmed by the person who knew him best.

---

Why This Matters

This isn’t magic. This is the preponderance hierarchy from the Genealogical Proof Standard.

The wife’s testimony on the death certificate carries weight because:

• Original source: The death certificate is an original record, created at the time of the event

• Primary information: The wife had direct personal knowledge of her husband’s name

• Direct evidence: The name stated directly answers the question “What was his full name?”

The marriage register wasn’t wrong—it just used a familiar form. The death certificate, with an informant who had decades of direct knowledge, provided the complete legal name.

I didn’t guess. I applied methodology.

---

What One Session Produced

The session summary—what I call the “gestalt”—shows the complete arc of one evening’s work. Phase one: marriage records from the 1860s through 1890s, establishing who married whom. Phase two: death certificates from the 1950s through 1970s, bridging forward to informants who remembered the earlier generations. Phase three: census records, filling gaps and confirming household compositions. But look beyond the numbers: this summary also identifies what we didn’t prove, what questions remain open, what sources we haven’t yet examined. The session doesn’t just celebrate discoveries—it maps the territory still unexplored. Tomorrow’s Day 5 will focus precisely there: the gaps.

Let me be specific about what emerged from this single session:

• 16 sources analyzed across three phases (marriage records, death certificates, census records)

• 22 persons documented with full GPS classification

• 3 name conflicts resolved using preponderance hierarchy

• Multiple family connections established with supporting evidence

The Joseph/Jethro Wilson resolution was one of three name conflicts we untangled:

• “Joseph” → “Jethro Wilson”: Resolved by wife’s testimony on death certificate

• “Alex” → “Henry Alexander”: The 1896 marriage register listed “Alex Lawrence,” but the 1955 death certificate named “Henry Alexander Lawrence”—”Alex” was a nickname

• “Rudolph” → “Reed”: The 1864 marriage bond listed “Rudolph Bare,” but later death certificates used “Reed”—the commonly used name form

Each resolution followed the same pattern: Claude noticed the discrepancy, I traced the sources, and the preponderance of evidence pointed to a single identity.

---

The CLAUDE.md That Made This Possible

Remember Day 3? The file that makes Claude yours?

This session used a CLAUDE.md containing the Genealogical Research Assistant v8—260 lines of standing instructions encoding GPS methodology:

• Anti-fabrication rules (never invent sources)

• Terminology guardrails (original/derivative, not primary/secondary for sources)

• Three-Layer Model definitions

• Preponderance hierarchy for conflict resolution

• Confidence level criteria

The methodology wasn’t improvised. It was specified. Claude didn’t discover GPS methodology during the session—Claude was constrained to follow it from the first prompt.

This is the architecture. The CLAUDE.md is the blueprint. The discovery is the result of methodology applied at scale.

---

Your Turn: The 5-Minute Task

You don’t need 140 images. You don’t need formal proof summaries. You need three files and one question.

Today’s task:

1. Add 3-5 genealogy files to your sandbox folder

2. Open Claude Desktop and navigate to your folder

3. Ask: “What do you see in these files?”

That’s it.

Claude will read the images. Extract names, dates, places. Note relationships. Report what it finds.

You might discover a name variant you hadn’t connected. A person appearing under different names in different records. A relationship implied but not stated. Or you might just get a clean summary of what your documents contain.

Either way, you’ve completed the cycle: Point. Ask. Discover.

---

What Comes Next

Tomorrow—Day 5—we look beyond individual discoveries to the larger picture: finding the stories and the gaps. The brick walls that define where your research ends and your next questions begin.

But today, make your first discovery. It doesn’t have to be a name resolution spanning fifty-four years. It just has to be yours.

---

May your patterns emerge, your discrepancies illuminate, and your discoveries feel earned.

---

Author: Steve Little prompting AI-Jane powered by Claude Opus 4.5

Read the series:

• Day 1: Meet Your New Research Partner

• Day 2: Your Safe Sandbox

• Day 3: The File That Makes Claude Yours

• Day 4: Your First Discovery (you are here)

• Day 5: Finding the Stories AND the Gaps (tomorrow)

From “Joseph” to “Jethro”: A Name Mystery Solved Today’s session demonstrates exactly this workflow in action. When I first examined the 1897 marriage register, it recorded “Joseph Little” marrying “Louler Bare.” A simple clerical entry—or was it? By correlating four independent sources spanning 54 years—the marriage register, a WWI draft card bearing his own signature, his death certificate, and his grave marker—we established that “Joseph,” “Jethro,” and “Joe” were all one man: Jethro Wilson Little. The clincher? His own hand, signing “Jethro Little” on a 1918 draft registration. A person knows his own name. Case closed.

From "Joseph" to "Jethro": A Name Mystery Solved

Today's session demonstrates exactly this workflow in action. When I first examined the 1897 marriage register, it recorded "Joseph Little" marrying "Louler Bare." A simple clerical entry—or was it? By correlating four independent sources spanning 54 years—the marriage register, a WWI draft card bearing his own signature, his death certificate, and his grave marker—we established that "Joseph," "Jethro," and "Joe" were all one man: Jethro Wilson Little. The clincher? His own hand, signing "Jethro Little" on a 1918 draft registration. A person knows his own name. Case closed.