Applied The Archive Playbook

Research Question and Search Log

Updated

Beginner

What you'll learn

~35 min
  • Ask an AI tool, in plan mode, for a report on your search log in a short, plain request
  • Answer its questions, and read its plan for the things that matter to the research
  • Log what each search covered: name forms, dates, places, and index or images
  • Read a search that found nothing as a statement about that search, kept with its cause
  • Keep a family story as a lead to test, not a finding

The story, and what you will end up with

Since January 2026 you have been looking for records of Great-uncle Walt, who served overseas and died in the war. On 8 February a relative tells you the family story about him, and you write it down as it is told:

“Great-uncle Walt flew the plane. It went down in China, in the mountains, carrying supplies. His name is on a memorial in India, and after the war he came home and is buried at Norrlid. Grandma always said he was lost on the twenty-second.”

Fictional training example.

The story is a lead: a set of things to test, not a set of findings. Some of it may hold up and some may not. You do not know yet, and you should not decide yet.

By then you have run four searches. Three found something. One found nothing. Without a log, you would soon forget which spellings you tried in the empty one, which years it covered, and whether you looked at an index or the page images. Without those, you cannot tell what the empty search rules out.

By the end of this lesson you will have a small program on your own computer, the report program, that reads your search log and prints a report: your question, the story as a lead, every search with what it covered, the records you know of, your next step, and a list of what is missing. You build it by asking your AI tool in plain English, and you check it by reading the report 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:

Terminal window
mkdir -p ~/archive-playbook/01-search-log && cd ~/archive-playbook/01-search-log

Start 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/01/search-log.json 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), the sample log is saved as data/search-log.json, and a first snapshot, your checkpoint, is saved. Later you can ask whether the file has changed since then. If it does not mention one of them, ask: “Did you save a checkpoint with the file in it?” If the download does not work, download search-log.json in your browser, check where it went (the browser’s list of downloads shows the folder), and tell your AI tool, for example: “search-log.json is in my Downloads folder. Copy that one file 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 the file 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 search-log.jsonclick to expand

One research question (Q-1), the family story as a lead (L-1), seven planned record sets (a record set is one kind of record you expect to search, such as a casualty file or a newspaper), four searches with what each one covered, the records you know of and how each can be reached, one next step, and a title for every record. Each item has a short code: Q for the question, L for a lead, RS for a record set, S for a search, R for a record.

REFERENCE DATA data/search-log.json downloaded starter search log open reference hide reference Reference example of the JSON file your agent downloads and reads. You do not type or edit this JSON by hand.
{
"sample": true,
"question": {"id": "Q-1", "text": "When and where did Sgt. Walter Olof Engstrom (born 1919, Kettle Ford, Tamarack County, Wisconsin) die, and where was he buried?", "period": "1944-06-20 to 1948-12-31", "localities": ["India", "Burma", "Wisconsin"]},
"leads": [
{"id": "L-1", "text": "Family story: he flew the plane; it went down in China; name on a memorial in India; buried at Norrlid; lost on the 22nd", "source": "R-31", "note": "told by a living relative; not yet tested"}
],
"planned": ["RS-1 draft and enlistment records", "RS-2 squadron morning reports, Jan to Apr 1945", "RS-3 loss report", "RS-4 casualty file", "RS-5 overseas burials register", "RS-6 Kettle Ford Courier, 1945", "RS-7 county honor roll"],
"searches": [
{"id": "S-01", "date": "2026-01-12", "set": "RS-1", "where": "enlistment index (online)", "names": ["Engstrom", "Engström", "Engstrum", "Ingstrom"], "dates": "1940 to 1946", "jurisdictions": "all states", "mode": "index only", "result": "found", "records": ["R-13", "R-14"]},
{"id": "S-02", "date": "2026-01-14", "set": "RS-5", "where": "overseas burials register (online index)", "names": ["Engstrom", "Engström", "Engstrum"], "dates": "1944 to 1950", "jurisdictions": "all theatres", "mode": "index only, no images", "result": "not found", "cause": "the register lists only burials that remained overseas; if his remains were returned home he would not be in it"},
{"id": "S-05", "date": "2026-02-03", "set": "RS-6", "where": "Kettle Ford Courier images", "names": ["Engstrom"], "dates": "1945-03 to 1945-08", "jurisdictions": "Tamarack County", "mode": "images, page by page", "result": "found", "records": ["R-23"]},
{"id": "S-06", "date": "2026-02-05", "set": "RS-7", "where": "county honor roll index", "names": ["Engstrom"], "dates": "all", "jurisdictions": "Tamarack County", "mode": "index only", "result": "found", "records": ["R-26"]}
],
"holdings": [
{"record": "R-12", "status": "online"}, {"record": "R-13", "status": "online"}, {"record": "R-23", "status": "online"}, {"record": "R-26", "status": "online"},
{"record": "R-24", "status": "request only", "held": "National Military Records Centre (fictional)", "note": "request submitted 2026-02-10"},
{"record": "R-18", "status": "request only", "held": "National Military Records Centre (fictional)", "note": "not yet requested"}
],
"next": [
{"action": "request R-18 morning reports, March 1945", "why": "R-26 gives 22 Mar 1945 with no source; the unit's own daily record is the nearer source"}
],
"records": {
"R-12": "Draft registration card, Walter Olof Engstrom",
"R-13": "Enlistment index entry, Walter O. Engstrom",
"R-14": "Enlistment index entry, Walter E. Engstrom",
"R-18": "Morning report, 23 March 1945",
"R-23": "Kettle Ford Courier, 17 May 1945",
"R-24": "Casualty file TRAINING-CF-2214 (12 pages)",
"R-26": "Tamarack County honor roll, online index (2009)",
"R-31": "Family story note, 2026"
}
}

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 who you are researching, which file to use, and what you want from it:

I'm researching my great-uncle Walt, who died in the war. Help me build a small tool that reports everything in my search log, data/search-log.json, including what's missing. Ask me what you need first.

Your AI tool reads the file and comes back with questions. It may also point out things it noticed in the log, such as the record sets nobody has searched yet. You do not have to act on those now.

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 asksYou might say
Is this real data, or practice? Should the program ever change the file?It’s the course’s practice file; read it and don’t change it.
Python, or something else?Use what’s already on this computer, and ask me before installing anything.
On screen, in a browser, or saved as a file?Print it in the terminal and save a copy I can keep.
How far should “what’s missing” go?Anything my log is missing: record sets I haven’t searched, and empty searches with no reason.
Should it judge the evidence, or check the family story?Don’t judge the evidence yet; just report what the log says.
Who is it for?Just me, for my own research.

Any language is fine, as long as it uses what your computer already has. Weighing the family story against the records is your work, and later lessons do it one detail at a time; this tool keeps the bookkeeping.

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:

  1. Every search shows what it covered: the spellings tried, the years, the places, and whether you searched an index (a typed list made from the records) or the page images. An empty search also keeps its reason, or says the reason is not known yet. An empty search means something only next to what it covered.
  2. The family story has its own heading, as a lead, and none of its details turn up anywhere as facts.
  3. Nothing says a record “does not exist”, “was lost”, or that a search “proves” anything. A search can only say what it did not find.
  4. Your file is read and left as it is. It is your log.
  5. The records you know of are listed, with how each can be reached, and your next step with its reason.
  6. What’s missing covers both the record sets with no search yet and any empty search with no reason given.

A first plan often leaves one of these out. Say so in a sentence, in your own words, for example:

Never say a record doesn't exist.

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. If it asks one last question first, answer it in the same message. Let it build. It may run what it built and tell you how to run it yourself; either way, the next step asks it to. 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 what it built:

Run it and show me the report.

Your report will be laid out in its own way. Check it for these:

  • The question Q-1 appears with its period, 1944-06-20 to 1948-12-31, and its three places: India, Burma and Wisconsin.
  • The family story L-1 is under its own heading as a lead, with its source, R-31.
  • S-02 shows as not found together with everything it covered: three spellings, 1944 to 1950, all theatres, index only, no images. Its cause is printed beside it.
  • Nothing in the report says a record does not exist.
  • The records you know of show how each can be reached: four online, and two request only (R-24, the casualty file, and R-18, a morning report), each with who holds it and its note.
  • The next step appears with its reason.
  • What’s missing includes RS-2, RS-3 and RS-4, the record sets with no search yet, and does not list S-02 as missing a reason, because it has one (whether that reason holds is still to be checked).
  • When you ask “Has my log file changed since the checkpoint?”, it compares the file with the checkpoint and says no.

The missing list may hold more than that. It may also list S-03 and S-04 as gaps in the numbering; those two searches are in a different log, about another branch of the family, which a later lesson uses. If something on the list is not true, tell your AI tool in a sentence what you see and what you expected. If the report leaves out the records and the next step, for example, say: “The report should also show the records I know of, how each can be reached, and my next step with its reason.”

Notice what the empty search can tell you

Find S-02 in the report, then find S-01. Before you read on, compare the spellings each one tried. What did S-02 leave out?

S-02 searched an overseas burials register, an online index with no page images, for three spellings, Engstrom, Engström and Engstrum, from 1944 to 1950, and found nothing. S-01 also tried Ingstrom; S-02 did not, and you can see that only because the log kept the spellings.

A hasty reader takes S-02 to mean “he was never buried overseas”, or “his record has been lost”. Neither follows. The note beside it gives one possible reason: the register lists only burials that remained overseas, so if his remains were returned home he would not be in it. You do not yet know whether that happened.

Now see what happens when that cause is missing. Ask:

What happens if S-02 has no reason? Try it on a copy in this folder, and leave my log alone.

The copy’s report should now list S-02 as an empty search with no reason given. If it does not, say “An empty search with no reason should count as something missing.” When you have looked, remove the copy: “Delete that copy; I’m done with it.” That gap is the one that gets misread later: a month from now, a bare “not found” looks like an answer.

Not found describes the stated search. It becomes evidence of absence only when the expected record and search coverage justify that interpretation.

Here, neither condition holds. You do not yet know whether Walt’s record is expected in this register (that depends on whether he was brought home), and the coverage is an index only. The step that can explain the empty result is the casualty file, R-24, which is already requested. A cause is your own sentence about why a record might not be there, a reason to check rather than something you have shown, and cause unknown; not yet assessed is an honest entry.

In the story it is 12 February 2026. What you can conclude then: the S-02 miss says what was searched, and the casualty file request is the step that can explain it.

One real run

Here is part of one report, built with Claude Code on 10 November 2026. Yours will be laid out differently; check it against the list above, not against this one.

## Leads (unverified, not evidence)
Leads are claims to test. They are not findings and are not counted as search results.
- **L-1**: Family story: he flew the plane; it went down in China; name on a memorial in India; buried at Norrlid; lost on the 22nd
- Source: R-31 (Family story note, 2026)
- Note: told by a living relative; not yet tested
- Status: lead only, not evidence
...
### S-02: overseas burials register (online index)
- Date searched: 2026-01-14
- Record set: RS-5 (overseas burials register)
- Where: overseas burials register (online index)
- Names tried: Engstrom / Engström / Engstrum
- Dates covered: 1944 to 1950
- Jurisdictions: all theatres
- Method: index only, no images
- Result: not found in overseas burials register (online index), index only, no images; names Engstrom / Engström / Engstrum; 1944 to 1950; all theatres. This covers only what was searched.
- Reason given: the register lists only burials that remained overseas; if his remains were returned home he would not be in it
...
## Gaps in the log
### Log structure
- none
### Planned record sets with no search logged
- RS-2: no search logged yet
- RS-3: no search logged yet
- RS-4: no search logged yet
...

Fictional training example.

This build labels the story “not evidence”. Read that as “not yet tested”: the story is a lead to check, not a finding.

How real archives differ

The repository in this case is fictional, and so is its request process. Real archives each have their own rules, so check their pages:

  • The US National Archives describes morning reports, a unit’s daily report of changes in each soldier’s status, as reports “produced every morning (hence their name) by an Army or Air Force clerk” (National Archives blog). For the archival ones it says: “Written requests for archival Morning Reports and Unit Rosters are not accepted.” Access is by “Visiting the NPRC Archival Research Room in St. Louis, MO” or by “Employing an independent researcher to examine the records on their behalf” (Morning Reports and Unit Rosters). So you cannot mail-order a real morning report the way this case’s repository allows. In your own log, mark one request only with the note in person or by a hired researcher; no mail requests.
  • For casualty files, the file of reports and letters about one service member’s death, start at the National Archives at St. Louis Frequently Requested Records page. It lists “Individual Deceased Personnel Files (IDPF)” for “Army, Navy, Marine Corps, Air Force: 1939-1975”, “Also known as Mortuary or Casualty Files”, and links a search to “Check which records are already on the Catalog”.

Keep going

Small changes do not need plan mode. Say each in a sentence, then run the report again:

  1. “Give me a way to add a new search.”
  2. “When a search found nothing, ask me why, and let me answer ‘not known yet’.”
  3. “Add a next step for RS-3, the loss report, and ask me why.”
  4. “Save a checkpoint.” Your AI tool saves a snapshot you can return to (people who use Git call this a commit). Some AI tools put it on a new branch (a separate line of snapshots) and say so; that is fine.

When you have your own searches to log, use the finished report 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:

  1. 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.
  2. While this folder still holds only the practice files, ask your AI tool: “How do I run this myself on a log in another folder, and save the report there too?” If it offers to copy your log into this folder instead, say no: your real log stays outside it.
  3. Write down what it tells you.
  4. Leave your AI tool: type /exit and press Enter (if you use Antigravity, press Ctrl+D with nothing typed instead).
  5. Keep your real files in a folder of their own, outside this course folder, and do what you wrote down. The 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 new search goes in your log: a ship passenger list index, arrivals at one port in one year, one spelling of the surname, and it found nothing. What would you ask your report program to record about that search, and what would you look at before reading anything into the empty result?

A good answer asks for the search’s coverage (one spelling, one port, one year, index only) and a reason or “not known yet”, and asks whether the record was expected in that list at all.

Key takeaways

  • Ask briefly, and let your AI tool ask back. Say who you are researching, which file to use and what you want from it; its questions fill in the rest.
  • Read the plan for the research, not the code: coverage beside every search, the story kept as a lead, no claim that a record does not exist, your file left alone.
  • Log coverage, not just results: name forms, dates, places and index or images decide what an empty search can mean.
  • An empty search keeps its cause: write why the record might not be there, or that you do not know yet, and read the result as a statement about that search.
  • Access is not existence: request only is a route to a known record, and the next step says why you are taking it.
KNOWLEDGE CHECK

A cousin searched an online death index for 'Pellander', 1950 to 1955, one county, index only, and found nothing. She writes: 'So Grandpa did not die in that county.' The family also spells the name Pellandar. What is the careful reading?

What’s next

In the next lesson you build the Source Register and Citations: a register of the files you collect, with where each came from, the date you got it and the page it shows, which also writes their citations.

intellimetrics · Arlington, VA
© 2026 intellimetrics.net
Search lessons