Claim Ledger
Updated
BeginnerWhat you'll learn
~40 min- Ask an AI tool, in plan mode, for a small program that shows your claim ledger (a numbered list of the claims you make from your records, each with the passage it quotes) and checks every quote
- Read its plan for what matters to the research: quotes checked character for character, your own judgments left as you wrote them, corrections added as new claims
- Read a quote that does not match as a fact about the quotation, and see where it parts from the transcription
- Keep a claim with a supporting and a contradicting record in view, without letting either one win by default
The story, and what you will end up with
You are working on the family line now: was the Olof Engstrom who married in Kettle Ford, Wisconsin, by 1886 the Olof born in 1858 in Gråbacka village, Tallmora parish, Sweden? On 21 February 2026 his 1931 death certificate (R-11) arrives by mail from the county, the copy you asked for. You now have two records side by side: that certificate, and the 1858 baptism entry (R-03) you compared line by line with the page image in the last lesson. The reviewed text of both is in one plain text file.
Your notes about Olof have grown into a list of sentences: “Olof was born on 14 March 1858.” “His parents were Erik Olofsson and Brita Persdotter.” You started some of them before you checked the baptism entry against the page. In a list of loose notes every sentence looks equally finished, and when you delete or overwrite one, nobody can see later that it was ever there.
So you keep them as a claim ledger. A claim is one sentence you believe a record tells you. Each claim has a number (F-01, F-02 and so on), says what it rests on (its basis: a record that says it, or other claims you worked it out from), and quotes the passage it comes from.
By the end of this lesson you will have a small program on your own computer that shows your ledger and checks every quote against the transcription, character for character. You build it by asking your AI tool in plain English, and you check it by reading what it prints.
Before you start
First, if your AI tool is still open, leave it: type /exit and press Enter (if you use Antigravity, press Ctrl+D with nothing typed instead), and wait until you see the $ or % prompt. The next line goes there, in the terminal, not in the AI tool. It makes a folder for this lesson and goes into it:
mkdir -p ~/archive-playbook/04-claim-ledger && cd ~/archive-playbook/04-claim-ledgerStart your AI tool with the start line you kept from the setup lesson, the one with approval prompts off. Then ask it to set the folder up:
Set this folder up so we can undo any change you make, download https://learn.intellimetrics.net/starters/history/04/ledger.jsonl and https://learn.intellimetrics.net/starters/history/04/transcriptions.txt into a data folder, and save a checkpoint.Look for it saying three things: the folder is set up with Git (the program that saves snapshots, from setup), both files are saved in data/ (ledger.jsonl is a text file with one entry per line), and a first snapshot, your checkpoint, is saved. If it does not mention one of them, ask: “Did you save a checkpoint with both files in it?” If the download does not work, download ledger.jsonl and transcriptions.txt in your browser, check where they went (the browser’s list of downloads shows the folder), and tell your AI tool, for example: “Both files are in my Downloads folder. Copy those two files into a data folder here, and save a checkpoint.” In Windows PowerShell or a desktop app, “my Downloads folder” works as is; only say “my Windows Downloads folder” if you use Ubuntu (WSL) on Windows, which keeps its own Downloads folder apart from Windows.
Everything in both files is fictional, made up for this course.
Open this box once, because it says what each code used further down stands for.
ℹWhat is in the two filesclick to expand
transcriptions.txt holds the reviewed text of two records, each under a header line: R-03, the baptism entry (seven lines of Swedish as written on the page, abbreviations kept; [?] marks a character nobody can read), and R-11, the death certificate. Each record ends with the line Fictional training example.
== R-03 ==Ao 1858. Födde och Döpte i Tallmora.N:o 9. Mart. 14 föd. 16 döpt.Olof.Föräldr: Bd. Erik Olofsson i Gråbackao. h. Brita Persdotter, 3[?] år.Fadd: dr. Hans Jonsson i Myrbo,pig. Kerstin Olsdotter i Gråbacka.Fictional training example.== R-11 ==Name: Olof EngstromDate of death: Feb. 17, 1931Place of death: Town of Kettle Ford, Tamarack Co.Born: March 14, 1857Age: 73 yrs 11 mos 3 daysBirthplace: SwedenFather: Eric OlsonMother: UnknownInformant: Eric Engstrom, Kettle FordI attended deceased from Feb. 10, 1931 to Feb. 17, 1931.Fictional training example.In English, R-03 reads: “Year 1858. Born and baptized in Tallmora. No. 9. March 14 born, 16 baptized. Olof. Parents: the farmer Erik Olofsson of Gråbacka and his wife Brita Persdotter, 3[?] years. Baptism witnesses: the farmhand Hans Jonsson of Myrbo, the maid Kerstin Olsdotter of Gråbacka.” In Brita’s line, an ink blot covers the second digit of her age.
This lesson names a line of the baptism entry by whose it is: Brita’s line (o. h. Brita Persdotter, 3[?] år.) or Hans Jonsson’s line (Fadd: dr. Hans Jonsson i Myrbo,). Your ledger program may give a line a number instead, and that number depends on where it starts counting: from the top of the file, or from the record’s own header line. So the same line can show as two different numbers, one apart. Either is fine.
ledger.jsonl is your ledger: a first line marking it as sample data, then one claim per line, in the order you wrote them. Each claim has its number, its sentence, its basis (Source statement when a record says it, Researcher inference when you worked it out from other claims, which it then names), and its links. A link names a record, quotes a passage from it, and gives your judgment of that passage: Supports, Contradicts, or Not assessed when you have not decided yet. A claim that corrects an earlier one says which one it replaces.
| Claim | What it says | Basis | Links (record: your judgment) |
|---|---|---|---|
| F-01 | Olof was baptized in Tallmora on 16 March 1858. | Source statement | R-03: Supports |
| F-02 | Olof was born on 14 March 1858. | Source statement | R-03: Supports; R-11: Contradicts |
| F-03 | Olof’s parents were Hans Jonsson and Kerstin Olsdotter. | Source statement | R-03: Supports |
| F-04 | Olof’s parents were Erik Olofsson and Brita Persdotter. | Source statement | R-03: Supports |
| F-05 | Hans Jonsson was a farmhand in 1858. | Source statement | R-03: Supports |
| F-06 | Hans Jonsson and Kerstin Olsdotter were baptism witnesses for Olof, not his parents. Replaces F-03. | Source statement | R-03: Supports |
| F-07 | Brita Persdotter was 31 at the baptism. | Researcher inference, from F-04 | R-03: Not assessed |
Switch to plan mode
Your AI tool is already open. Switch to plan mode now: Shift+Tab until it says plan, or /plan (the keys for each tool). In plan mode it usually reads your file, asks its questions and shows you a plan before it builds anything; the order can vary.
The conversation
Each message below is an example. Say it your own way.
Say what you want
Say whose claims they are, which files to use, and what you want:
My claims about Olof are in data/ledger.jsonl, each quoting a passage from data/transcriptions.txt. Help me build a small program that shows my ledger and checks every quote against the transcriptions. Ask me what you need first.Your AI tool reads both files and comes back with questions. It may already tell you which quotes it could not find in the transcription; you will read them in the program’s own output once it is built.
Answer its questions
Answer each in a sentence. These are questions AI tools often ask, with one way to answer each. Yours may ask them in a different order or show a list of choices:
| If it asks | You might say |
|---|---|
| How exact must a quote be? | Exactly, character for character; show me where a quote differs. |
| Check a quote against the whole file, or only the record the link names? | Only the record the link names. |
| Can a quote run over more than one line? | No, keep each quote to one line. |
| How should a replaced claim, like F-03, appear? | Show it, marked as replaced by F-06. |
| Should it also check the claim’s own wording against the quote? | No, only the quotes. Whether a passage supports a claim is my call. |
| In the terminal, or a page in the browser? | A report in the terminal is fine; use what’s already on this computer. |
| Should it ever change the ledger? | It can add new claims at the end, but never change or delete one. |
If it asks several questions at once, answer the ones that matter to your research in a sentence, and for the rest say that its suggestions are fine. If it asks something not listed here, answer it the same way, in a sentence.
Read the plan
Next it shows you a plan: what it will build and how it will check it. Read it for six things:
- Every quote is checked character for character against the record its link names, with no tidying of spelling, spaces or abbreviations. A quote that does not match shows where it parts from the line it came closest to, and nothing corrects it for you.
- Your judgments stay yours. Supports, Contradicts and Not assessed are shown as you wrote them, and nothing calls a claim verified, confirmed or proven.
- A claim with a supporting and a contradicting link shows both, side by side.
- Corrections are new claims added at the end. A replaced claim stays in the list, marked with the claim that replaced it. Nothing in the ledger is edited or deleted.
- A claim you inferred shows which claims it rests on.
- Your transcriptions file is read and left as it is.
A first plan often leaves one of these out. Say so in a sentence, in your own words, for example:
Never call a claim verified or proven; whether it's supported is my call.When the plan has all six, approve it: if your AI tool offers a choice to go ahead, pick it; otherwise say “Yes, go ahead and build it.” In Codex, press Shift+Tab first to leave plan mode. In Copilot, saying it brings up a choice to go ahead; pick that. Let it build. It may test what it built and tell you how to run it. When it has finished, ask it: “Did you install anything, and where?” Nothing, or something inside this lesson’s folder, is fine for these lessons. If it names anywhere else on your computer, ask it what that is for, and to remove it if you do not want it there.
Look at what it built
Ask it to run the program:
Run it and show me my ledger.Your ledger will be laid out in its own way. Check it for these:
- The seven claims carry eight quotes, because F-02 quotes both records. Six match. Two do not: F-03’s and F-05’s. For each, the output shows where the quote parts from Hans Jonsson’s line in the baptism entry: F-03 has
Fader: Dr.where the transcription hasFadd: dr., and F-05 hasdrängwhere it hasdr. - F-02 shows both its links, each with its quote: R-03 as Supports and R-11 as Contradicts.
- F-03 is still listed, with its claim and its quote, marked as replaced by F-06.
- F-07 shows that it rests on F-04, that its quote matches, and that its judgment is still Not assessed.
- No line calls a claim verified, confirmed or proven.
- When you ask “Have my two files changed since the checkpoint?”, it compares them with the checkpoint and says no.
If something on the list is not true, tell your AI tool in a sentence what you see and what you expected, for example: “F-03 is missing from the list. Show replaced claims too, marked with what replaced them.”
Notice what a matching quote can tell you
Start with the two quotes that do not match. F-03 quotes Hans Jonsson’s line as Fader: Dr., the reading you corrected in the last lesson: the word is Fadd:, short for faddrar, baptism witnesses, and not Fader, father. F-06 already corrects that claim. F-05 is different: its claim, that Hans Jonsson was a farmhand, is your reading of dr. as dräng, farmhand, and it may well be right. But someone wrote dräng inside the quotation marks. A quotation keeps the characters on the page; your reading of them belongs in the claim, where anyone can see it is yours.
Now F-02, “Olof was born on 14 March 1858”. The baptism entry supports it: N:o 9. Mart. 14 föd. 16 döpt., born on the 14th of March, in the entry for 1858. The death certificate contradicts it: Born: March 14, 1857. Both quotes match. A hasty reader drops the certificate because it disagrees, or lets it win because it is an official document. Look at who gave the certificate its birth date: the informant, the person who supplied the details, was “Eric Engstrom, Kettle Ford”. The certificate does not say who he was. If he was Olof’s son, born long after Olof, he could not have known the birth date first hand. Weighing the two records is the next lesson’s work. The ledger’s job is to keep both in view until then.
Last, F-07: “Brita Persdotter was 31 at the baptism.” Its quote matches. Before you read on, look at what the quote holds.
It holds 3[?]: a 3, then a digit nobody can read. See what the check does with a different age. Ask:
In this folder, on a copy of my ledger, add a claim that Brita was 39 at the baptism, quoting her line exactly. What does the check say?The quote matches again. The check found the same characters for 39 as for 31, because it compares the quote with the transcription, not the quote with the claim. When you have read the answer, say “Delete the copy.”
An exact quotation shows what this text says. The researcher must decide whether it supports the particular claim, and with what limits.
Here nothing supplies the second digit. F-07 names F-04 as what it rests on, and F-04 says who Brita was, not how old. The 31 came from the AI draft of the transcription, before your own line-by-line check of it against the page in the last lesson put 3[?] in its place.
Correct F-07 without losing it
Replace F-07 with a claim the passage can carry:
Now correct F-07 in my real ledger: add a claim replacing it that says her age reads 3[?], second digit unread.Then ask “What has changed in my two files since the first checkpoint?” The answer should be one new claim, F-08, added at the end of the ledger, replacing F-07; F-07 itself unchanged and still there; and the transcriptions file unchanged. Your AI tool may save a checkpoint on its own, and may choose a basis or a judgment for F-08 and tell you which; if you want a different one, say so, and it adds another claim that replaces F-08.
One real run
Here is part of the output one program printed, built with Claude Code on 10 November 2026. Yours will be laid out differently; check it against the list above, not against this one.
F-05 claim: Hans Jonsson was a farmhand in 1858. basis: Source statement R-03 · Supports quote: Fadd: dräng Hans Jonsson i Myrbo result: Quote differs from record text closest line in R-03 (transcriptions line 7): quote: Fadd: dräng Hans Jonsson i Myrbo ^^^ record: Fadd: dr. Hans Jonsson i Myrbo, ^ differences: - quote has "äng", record has "."
...
F-07 claim: Brita Persdotter was 31 at the baptism. basis: Researcher inference based on: F-04 R-03 · Not assessed quote: o. h. Brita Persdotter, 3[?] år. result: Quote matches record text (transcriptions line 6, column 1)Fictional training example.
This program counts lines from the top of the file, so Hans Jonsson’s line shows as line 7 and Brita’s as line 6. “Quote matches record text” is all the check can say about F-07; the 31 is still yours to judge.
How real archives differ
The register in this case is invented, and so are its abbreviations. Some of what the lesson leans on comes from real sources:
- The Swedish National Archives (Riksarkivet) says, in Swedish, that birth and baptism books give “namn på faddrar eller dopvittnen” (the names of godparents or baptism witnesses) (Riksarkivet, kyrkoarkiv; the Swedish words are the archive’s own, the English in brackets is a translation). So faddrar, the word behind
Fadd:, is a real word from the archive; the short formFadd:belongs to the register made up for this course. - “First hand” has a name in the method many genealogists follow. Elizabeth Shown Mills calls “information based on firsthand knowledge” primary and “information based on secondhand knowledge” secondary, and adds a third value: “Without that identification, the quality of our information remains undetermined” (Evidence Explained, QuickLesson 17). The death certificate’s birth date is a candidate for secondary until you know who Eric was.
- The Board for Certification of Genealogists lists “Resolution of conflicting evidence” as one of the five components of its Genealogical Proof Standard (BCG, Ethics and Standards). Keeping F-02’s two links side by side is the start of that work, not the end of it.
Keep going
Small changes do not need plan mode. Say each in a sentence, then look at the ledger again:
- “Correct F-05: add a claim replacing it, with the quote copied exactly from the transcription.”
- “F-06 quotes only Hans’s line. Add a claim replacing it that also links Kerstin’s line and the parents’ line.”
- “Add a note under each quote saying what it can’t prove.”
- “Save a checkpoint.” Your AI tool saves a snapshot you can return to (people who use Git call this a commit).
When you have claims and transcriptions of your own, use the finished program. It runs on your real records by itself, with your AI tool closed. Do these steps in order; the first two happen now, while this folder still holds only practice data:
- Go through the data check, the questions in setup about what the material is, who decides how it is used, and which AI tool would see it. Go on only if it allows this use.
- While this folder still holds only the practice files, ask your AI tool: “How do I run my ledger program myself on files in another folder, and save its report there too?”
- Write down what it tells you.
- Leave your AI tool: type
/exitand press Enter (if you use Antigravity, press Ctrl+D with nothing typed instead). - Keep your real files in a folder of their own, outside this course folder, and do what you wrote down. Its report is saved beside your real files, not in the course folder.
Try another case
Before the quiz, write two or three sentences for yourself. A cousin’s notes say “Grandma Anna was a dressmaker in 1910”, quoting a census line as Anna, dressmaker. When you open the census, the line reads Anna, dress mkr. What would you put in the quotation, and where would your reading of mkr. go?
A good answer quotes Anna, dress mkr. as the page has it and puts “dressmaker” in the claim as the cousin’s reading.
Key takeaways
- Ask briefly, and let your AI tool ask back. Say whose claims they are, which files to use and what you want; its questions fill in the rest.
- A quotation is the characters on the page: no expanded abbreviations, no tidied spelling. Your reading goes in the claim.
- A matching quote is not a supported claim: the check finds the characters; whether they say what your claim says, and with what limits, is your decision.
- Keep the contradiction in view: a claim can hold a supporting and a contradicting record at once, and neither wins by being dropped.
- Correct by adding, never by overwriting: the replaced claim stays, marked with the claim that replaced it.
Your ledger holds the claim 'Anna Pellander kept a shop in 1902', linked to a city directory line quoted as 'Pellander A., grocer, Mill St.' The quote check says it matches. The same directory also lists a second A. Pellander on another street. What is the careful reading?
What’s next
In the next lesson you build the Evidence Weigher: a way to lay out every source of a disputed fact, such as the date Great-uncle Walt (Walter Olof Engstrom) was lost, by whether it is an original or a copy, whether its informant knew the fact first hand, and how the account reached the page, so that copies of one account count once. You write the conclusion and what still contradicts it.