Then-and-Now Comparer and Feature Register
Updated
IntermediateWhat you'll learn
~50 min- Ask an AI tool, in plan mode, for a page that compares two aerial photographs and keeps your notes on what each one shows
- Read its plan for notes kept per exposure (one photograph, taken at one moment), never per scan (one digital copy of it), with every feature on your list kept and nothing filled in for you
- Switch between two pictures to find what differs, and say which picture each difference is on
- Record for each feature and each exposure one of five words: present (you can see it), clearly absent (the ground shows and it is not there), obscured (something hides the ground), outside coverage (the photograph does not reach that far) or not assessed (you have not looked yet), telling a feature hidden by cloud from one that is gone, and ground outside the photograph from empty ground
- Keep a mark that is on one scan only in a separate list, tied to that scan
The story, and what you will end up with
Most of Walt’s story is in records: registers, letters, reports. A few of the family’s sources are pictures, and two of them are aerial survey photographs of the home county, taken from an aircraft looking straight down at the Engstrom farm near Kettle Ford, Tamarack County, Wisconsin, where Walt grew up. Exposure A, frame 0412, was taken on 4 February 1945. Exposure B, frame 0077, was taken on 18 April 1945. In February 2026 you obtained scans of both from the Air Photo Library (fictional), the archive in this case that holds the film. You have one scan of B and two of A, made a week apart: two copies of the same moment. They are the three aerial scans you listed in the photograph and scan catalog, where your notes already say that the second scan of A carries a dust speck.
Nothing in the case connects these photographs with Walt’s loss or burial: he was lost over Burma, and these show the family’s farm. They are here to practise reading two images, not to place an event.
Put A beside B and things differ. Something has changed where the farm track meets the creek. A pale round patch appears. A white shape lies over the ridge, and the right-hand edge of B is a band of grey stripes. It is very easy, at this point, to start telling a story about what happened in the ten weeks between them.
By the end of this lesson you will have a page on your own computer that shows the three scans, lets you switch between any two of them, and keeps your notes: for each feature on your list and each exposure, what you can see, and only that. You build it by asking your AI tool in plain English, and then you look at the pictures and fill in the notes yourself.
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/11-comparer && cd ~/archive-playbook/11-comparerStart 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/11/aerial-frames.zip and unpack it 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), three pictures and a file called frames.json (a text file of data about the pictures) are in the data folder, and a first snapshot, your checkpoint, is saved. If it does not mention one of them, ask: “Did you save a checkpoint with the files in it?” If it says it kept the zip file somewhere outside this folder, ask it to delete that copy. If the download does not work, download aerial-frames.zip in your browser, check where it went (the browser’s list of downloads shows the folder), and tell your AI tool, for example: “aerial-frames.zip is in my Downloads folder. Unpack 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 download is invented for this course: the farm, the photographs, the archive and the dates. The pictures are drawings made for this course.
Open this box once, because it says what each code used further down stands for.
ℹWhat is in aerial-frames.zipclick to expand
Three drawn pictures in an images folder, one per scan, and frames.json, your notes on them: the place, the two exposures, the three scans with the dates they were made, and your list of eight features to look for.
| Scan | Exposure | Taken | Scanned |
|---|---|---|---|
TRAINING-AER-A-0412-S1 | A, frame 0412 | 1945-02-04 | 2026-02-02 |
TRAINING-AER-A-0412-S2 | A, frame 0412 | 1945-02-04 | 2026-02-09 |
TRAINING-AER-B-0077-S1 | B, frame 0077 | 1945-04-18 | 2026-02-09 |
A scan’s name is its exposure’s name with -S1 or -S2 (scan 1, scan 2) added. Each picture carries its exposure’s name and date in the top left corner and “Fictional training example.” along the foot.
A place on a picture is given as two numbers, x and y: x counts from the left edge, y from the top edge. Every picture is 1000 across and 700 down, and the three are already lined up, so (430, 385) is the same spot of ground on all three.
| Code | Feature | Where to look |
|---|---|---|
| RIV | creek | runs west to east across the picture |
| FB | footbridge | where the farm track meets the creek, near (430, 385) |
| VIL | farm buildings (six) | six squares, near (600 to 685, 215 to 270) |
| TRK | farm track | from (430, 365) to (600, 240) |
| CLR | clearing | near (300, 230) |
| HUT | ridge shed | on the ridge line, near (210, 100) |
| FLD | east field | a rectangle near (880 to 960, 260 to 330) |
| SPK | object at 512, 455 | near (512, 455) |
How to read the drawings: white is cloud, as it is in a photograph taken from above, and a band of grey diagonal stripes is ground beyond the edge of the photograph, which it does not show.
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 photographs they are, where they are, and what you want to do with them:
Two aerial photographs of Walt's family farm, taken in February and April 1945, are in the data folder. Help me build a page to compare them and write down what each one shows. Ask me what you need first.Your AI tool looks at the pictures and comes back with questions. It may also tell you what it noticed: that there are three scans, not two, that a dot is on one February scan only, or that something has gone from the April picture. Treat what it says as a first look, not as your notes; you look for yourself once the page 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 |
|---|---|
| Is this practice material, or real photographs of the farm? | It’s the course’s practice material. Read the files and don’t change them. |
| Which February scan should the page use? | Show all three scans so I can compare any two. |
| Should I label the dot on scan 2 as probably dust? | Don’t label the dot; I’ll decide what it is. |
| Side by side, a slider that wipes from one to the other, or flicking between them? | Let me flick between them, and side by side too. |
| What do you want to look at: the whole picture, or particular things? | Every feature on my list, for each photograph. |
| Should I write the notes for you, or leave them blank for you to fill in? | Leave every feature on my list blank for me to fill in. |
| Where should the notes be kept: in the page, or in a file? Do you need to download them? | Save my notes to a file here. |
| Should the notes say only what is visible, and not why it changed? | Yes, only what’s visible. |
| Do you know anything about the pictures from elsewhere, such as what the striped band is? | White is cloud; the striped band is ground the April photograph doesn’t cover. |
| Do you know anything about these changes from elsewhere, such as a flood? | No, nothing beyond the photographs. |
| Is it just for you, or for others too? | Just me, on this computer. |
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. Plans can be written in technical words, so for each of the six things below there is what it looks like in a plan, and a question to ask your AI tool if the plan does not say:
- All three scans can be shown, and you can switch between any two, the two February scans included. In a plan: all three scan names, or a choice of any two scans. Ask: “Can I switch between the two February scans?”
- Your notes are kept per exposure: one set for the February photograph and one for the April photograph, never one per scan. Both February scans share one set. In a plan: notes named by photograph or date (February and April, or A and B), two columns, sixteen boxes. A box, column or field for each scan or each image file, or three columns, means notes per scan. Ask: “How many sets of notes will there be, and is each one for a photograph or for a scan?”
- Every feature on your list is in both sets, the object at (512, 455) too, and every note starts empty. Nothing is filled in for you, and nothing says what changed or why. In a plan: all eight features named, or “every feature”, and words such as empty or blank. Ask: “Which features will be in my notes, and will any note be filled in before I write it?”
- A mark that is on one scan only has a list of its own, tied to that scan, apart from your notes. In a plan: a separate list or section for marks, with the scan, the place and what you see. Ask: “Where do I write down a mark that is on one scan only?”
- You get your notes as a file in this folder. In a plan: a file name, such as
notes.json, in this folder. Ask: “Where will my notes be saved?” - Your pictures and
frames.jsonare read and left as they are. In a plan: words such as read only, or not changed. Ask: “Will you change my pictures or frames.json?”
A first plan often keeps one set of notes per scan, three columns instead of two, and has no place for marks on one scan. Say so in a sentence, in your own words, for example:
Keep notes per photograph, not per scan, and list marks on one scan separately.After a change like that, read the plan again. If it has now decided by itself what the dot is and taken it off your list, say: “Keep every feature on my list in the notes, the dot too. I’ll decide what it is.”
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 start 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 start the page:
Start the comparison page and open it in my browser.Your AI tool may open the page straight from its file, or start a small program on this computer that serves the page at a local address, one ending in a number, such as http://127.0.0.1:8000/ (127.0.0.1 is an address that means this computer); either is fine. If it gives you an address without opening it, copy the address into your browser’s address bar and press Enter. If it says the address is already in use, say: “Something else is using that address. Don’t stop it. Use a different address for my page.” Leave the program running while you work; when you are done, ask your AI tool to stop it. If your AI tool does not take a message after opening the page, wait a minute. If it still does not, close the terminal window, open a new one, type this lesson’s folder line again, start your tool, and say: “Start the page again so it keeps running while we talk.” The page and your saved work are files in the folder, so nothing is lost.
Your page will be laid out in its own way. Find three things on it: where you choose which two scans to compare, where you switch between them, and where your notes go. If you cannot find one of them, tell your AI tool what you see and ask it where that is.
Look first, then write
Every note starts empty, and an empty note means you have not looked yet. When you write a note, start it with one of five words, so that each one says exactly what you saw:
- Present: you can see the feature.
- Clearly absent: you can see the ground there, clearly enough that the feature would show if it were there, and it is not there.
- Obscured: something covers the ground there (cloud, shadow, trees, a fold in the print), so you cannot see whether the feature is there.
- Outside coverage: the photograph does not reach that far.
- Not assessed: you have not looked yet. A blank note says the same.
A note may go on after its word, as in “Obscured, under the cloud”. Save as your page lets you: there may be a save button, or it may save as you type. If saving downloads a file instead, find its name in the browser’s list of downloads and tell your AI tool, for example: “My notes, notes.json, are in my Downloads folder. Move that one file into this folder.”
Here is how the pictures work out, feature by feature, with the reason for each note.
-
Switch between the two exposures. Compare February scan 1 (
TRAINING-AER-A-0412-S1) with April (TRAINING-AER-B-0077-S1), switching back and forth a few times. Four things jump: something at the creek near (430, 385), a pale circle near (300, 230), a white shape over the ridge, and the stripes on the right. Look at each one on its own before you write anything. -
The footbridge (FB): a change you can see. In February a short dark bar crosses the creek where the farm track comes down to it. In April the creek runs straight through that spot, and the ground there is plain to see. Write
Presentfor February andClearly absentfor April. A hasty reader goes on to write why it went. Nothing in either picture says, and neither do your notes: they say the bridge is visible in February and not in April. -
The clearing (CLR): a change the other way. Near (300, 230), April has a pale round patch that February does not. Write
Clearly absentfor February andPresentfor April. “Clearing” is your list’s name for a pale round patch;Presentsays the patch is there, not that your name for it is right. -
The ridge shed (HUT), under the cloud. In February a small dark square sits on the ridge line near (210, 100). In April that stretch of ridge is covered by the white cloud. The hasty reading is “the shed is gone by April”. But in April you cannot see the ground there at all. Write
Presentfor February andObscured, under the cloudfor April. The shed may still be standing under that cloud, or it may not; the April photograph cannot tell you. -
The east field (FLD), beyond the edge. In February the pale rectangle near (880 to 960, 260 to 330) is the east field. In April that part of the picture is grey stripes: the April photograph stops at x = 820. The hasty reading is “the field has gone”. The careful one is that the April photograph does not cover that ground. Write
Presentfor February andOutside coveragefor April. -
The dot (SPK), on one scan only. Now compare February scan 1 with February scan 2, switching between them. They are the same except for one small black dot at about (512, 455), below the creek and to the right of the footbridge, on scan 2 only. If you cannot see it, zoom in with your browser (Ctrl and +, or Cmd and + on a Mac) and watch that spot as you switch.
What you see: a dark dot on scan 2, and nothing at that spot on scan 1.
Why it cannot be on the ground: both scans were made from one negative (the original film a print or scan is made from), so they show one moment. Something on the ground would be on both. A dot on one scan of the exposure and not the other came from the scanning, usually dust, a hair or a scratch. Your catalog already notes a dust speck on scan 2, but you do not need to know the cause to know the dot is not on the ground.
Where it is written down: in your list of marks on one scan, as scan
TRAINING-AER-A-0412-S2, at 512, 455, “dark dot, on this scan only”. Your notes still have the SPK row, like every feature on your list, and it says what the ground shows at that spot: for February,Clearly absent, because scan 1 shows the spot plainly with nothing on it, with a note such as “dot on scan 2 only, in the marks list”; for April,Clearly absent.What it is not written down as: an object on the ground,
Presentfor February, or a sighting on scan 2 that counts again because there are two scans. -
The features that did not change. The creek (RIV), the six farm buildings (VIL) and the farm track (TRK) are visible in both: write
Presentsix times. In April the creek stops at the striped band, but it is plainly there up to it, so it isPresent.
Your sixteen notes now read like this:
| Feature | February, exposure A | April, exposure B |
|---|---|---|
| RIV creek | Present | Present |
| FB footbridge | Present | Clearly absent |
| VIL farm buildings | Present | Present |
| TRK farm track | Present | Present |
| CLR clearing | Clearly absent | Present |
| HUT ridge shed | Present | Obscured, under the cloud |
| FLD east field | Present | Outside coverage |
| SPK object at 512, 455 | Clearly absent (scan 1 shows nothing; the dot on scan 2 is in the marks list) | Clearly absent |
Two changes you can see, one feature hidden, one beyond the edge, and a mark on one scan kept in its own list. You chose every one of those words; the page chose none.
Check it for these
- The page shows all three scans by name, each with the date its exposure was taken (1945-02-04 for both February scans, 1945-04-18 for April) and the date it was scanned.
- When you ask “How are my notes stored: one set per photograph or one per scan, and which features are in each?”, it says two sets, one for each exposure, with all eight features in each: sixteen notes.
- Switching between February scan 1 and February scan 2 shows one difference only, the dot near (512, 455) on scan 2.
- In April, nothing but grey stripes lies right of x = 820: the creek stops at the band, and there is no east field.
- After you save and reload the page, your sixteen notes are still there, and the dot is in the list of marks under
TRAINING-AER-A-0412-S2only. - When you ask “Have my picture 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: “My notes have a column for each scan. Keep one set per photograph.”
Notice what a list of differences cannot tell you
Your page shows differences and leaves the deciding to you. This next step is an experiment: you ask for a small program that only compares the pictures, dot by dot, to see what such a program can and cannot tell you, and you remove it afterwards. Ask:
In this folder, write a small program that finds every place two pictures differ. Run it on February scan 1 against April, and on the two February scans. Leave my notes alone.Read what it reports, then put it beside your notes. Between February scan 1 and April it should find the striped band, the white cloud, the pale patch and the place where the footbridge was, and perhaps the printed date in the corner and a few tiny places along the edges of shapes. It finds no shed and no east field on their own: they are inside the cloud and inside the band, and a program comparing colours has no way to say whether the shed is still under that cloud. Between the two February scans it finds exactly one place, the dot, reported just as the footbridge was: a place where the pictures differ.
Every place on that list is a difference between two pictures. Your notes say what each difference is: a change on the ground, ground hidden from view, ground the photograph does not reach, or a mark on one copy. Then remove the program and anything it made, such as a list of results:
Remove that program and anything it made, stop the comparison page, and save a checkpoint.Obscured is not absent. Two scans of one exposure are not two observations.
One real run
Here is the list one program gave, written by Claude Code on 10 November 2026 when asked for the message above. Yours will be laid out differently and may count the places differently; check it against what this lesson describes, not against this one. A pixel is one of the tiny dots a picture on screen is made of; the last line says that 61 of them, around (512, 455), are the pale ground colour on scan 1 and nearly black on scan 2.
February scan 1 vs April: 6 places
| # | Where (x, y) | What changes in the picture |
|---|---|---|
| 1 | 820-999, 0-699 | The right-hand strip becomes the grey striped band (the ground April doesn’t cover). |
| 2 | 149-205, 7-19 | Printed caption at the top: frame number. |
| 3 | 219-390, 7-23 | Printed caption at the top: exposure date. |
| 4 | 110-330, 50-150 | A white area, which you identified as cloud, appears over the ridge line. |
| 5 | 278-322, 208-252 | A lighter round patch appears around (300, 230). |
| 6 | 425-434, 365-404 | The dark bar around (430, 385) in February isn’t there in April. |
February scan 1 vs February scan 2: 1 place
- x 508-516, y 451-459, centred on (512, 455): 61 pixels go from the background colour to near-black in scan 2. Nothing else differs between the two February scans.
Fictional training example.
Place 4 covers the ridge shed without naming it, and place 1 covers the east field.
How real archives differ
Real aerial photographs come with their own finding aids and their own terms, and each holder says in its own words how its images may be used. Read the current terms yourself before you use or share any image.
- Real frames are found by the exposure, not by your file. For one large series, the US National Archives says: “This photography can be found in the series Aerial Photographs, 1935 - 1970, from Record Group 373, Records of the Defense Intelligence Agency”, and “To locate the photographs, you need a citation with the spot number and exposure number.” Film is seen in person: “Once you order the cans at our facility, you can view and copy them in the Cartographic Research Room.” It adds: “NARA has a partnership with the National Collection of Aerial Photography, Scotland. The photographs scanned by NCAP are currently being added to the National Archives Catalog” (RG 373 aerial photography).
- USGS EarthExplorer holds “a large and diverse group of imagery acquired by Federal organizations from 1937 to the present” (USGS aerial single frames), and “EarthExplorer can be used to search, preview, and download existing medium resolution digital files at no cost.” USGS states that “USGS-authored or produced data and information are considered to be in the U.S. Public Domain” and asks “that proper credit be given” (USGS copyrights and credits). Read that carefully: the frames came from several federal organizations, and the public-domain sentence speaks of what USGS itself made.
- The Library of Congress says, for its prints and photographs: “In all cases, it is the researcher’s obligation to determine and satisfy copyright or other use restrictions when publishing or otherwise distributing materials found in the Library’s collections” (rights research guide).
- Real photographs do not arrive lined up. The practice pictures share one grid, so a point in A sits on the same point in B. Two real photographs taken weeks apart, from different heights and angles, do not. Fitting an old photograph to a map or to another photograph, called georeferencing, is a separate skill this course does not cover. Switching between two photographs that are not lined up shows every shift of position as a change.
Comparing photographs is desk work. This course gives no advice about visiting, searching or disturbing any place, and a place connected with a wartime loss may be a grave.
Keep going
Small changes do not need plan mode. Say each in a sentence, then look at the page again:
- “Add a magnifier that shows the same small patch of both pictures, enlarged, side by side.” It should enlarge only; if it sharpens or changes either picture, ask it not to.
- “For each February note, let me tick which scans I looked at.” The ticks stay inside the one February note; if it adds a column per scan, say: “Keep one set of notes per photograph.”
- “Give me a plain-text copy of my notes and marks I can print, with both exposure dates.”
- “Save a checkpoint.” Your AI tool saves a snapshot you can return to (people who use Git call this a commit).
This page compares only its three practice pictures, which come already lined up; real photographs have to be lined up first. When you have real photographs of your own, keep notes in the same shape: one row per feature, one column per exposure, the five words, and a separate list for marks on one scan. Then use the page on them. 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 pictures, ask your AI tool: “How do I start this page myself on pictures in another folder, with the notes saved beside them?”
- Write down what it tells you.
- Ask your AI tool to stop the page’s program, then leave the tool: type
/exitand press Enter (if you use Antigravity, press Ctrl+D with nothing typed instead). - Keep your real pictures in a folder of their own, outside this course folder, and do what you wrote down. The notes are saved beside your pictures, not in the course folder.
Try another case
Before the quiz, write two or three sentences for yourself. A cousin sends two scans of a 1938 aerial photograph of her grandparents’ farm and one of 1951. Both 1938 scans show an orchard east of the house. The 1951 photograph stops partway across the farm, and the ground where the orchard stood lies beyond its edge. She writes: “Orchard seen twice in 1938, gone by 1951.” What would you write for the orchard in 1938 and in 1951?
A good answer writes Present once for 1938, in one note, because two scans of one exposure show one moment, and Outside coverage for 1951, not Clearly absent, because that photograph does not show the ground there at all.
Key takeaways
- Ask briefly, and let your AI tool ask back. Say whose photographs they are, where they are, and what you want; its questions fill in the rest.
- Read the plan for the research, not the code: one set of notes per exposure, every feature on your list kept, nothing filled in for you, marks on one scan listed apart, your pictures left alone.
- Write down what you can see, per exposure, and nothing more: present, clearly absent, obscured, outside coverage, or not assessed yet. Beyond the edge of a photograph the word is
Outside coverage, notClearly absent. - Count exposures, not files: a mark on one scan belongs to that scan’s file, not to the ground.
- A change you can see is not a cause you can state: your notes say the footbridge is clearly absent in April. Why is a question for the records.
You are comparing a different pair: two scans of a 1952 exposure of a farm, and a 1958 exposure of the same farm. A small shed by the gate is clear on both 1952 scans. In 1958 a dark tree shadow falls across the gate and you cannot make out the ground there. How do you fill the shed's row?
What’s next
Your notes now say what each exposure shows and no more. The last lesson puts the whole case together: in The Case File: Summary Review you check a flawed, AI-written summary of Walt’s story sentence by sentence against the records it cites, and decide yourself, before anything is shared, whether each passage supports its sentence.