AI makes it possible to write interesting and useful programs even if you don’t know how to code. Describe what you want, wait a few minutes, and wow! Here’s your site.
It’s called “vibe coding.”
It’s not quite that easy. The first draft appears in a minute or two. It might need some tweaking to work. And then, as you see the possibilities, you add more functionality. It’s easy to spend hours in a back-and-forth conversation.
Here are several vibe-coded projects that tell stories of Little Compton history, or more generally, tools for local history and genalogy research.
Working with Claude Code is interesting. You tell it what you want and it tries to produce it. Claude brings a lot of knowledge to the partnership, beyond coding; partnership seems a good way to think of the relationship. A very patient partnership; you’ll spend a lot of time telling Claude what doesn’t work, and Claude will spend a lot of time explaining to you how to do things: upload this here, type this here. (You don’t need to know much about computers to do this work; any questions, ask Claude.) Claude also carefully explains what it’s doing as it works on the project, especially when it responds to problems you point out, which I found fascinating. Each of these projects represents a few hours of back-and-forth. In part, that’s because I kept asking for more features.
None of these are quite ready for museum public use. Before making them public, I would want to tweak them a bit, add a few features, and check the underlying data. They’re proof of concept.
Claude costs $20/month for the tier with coding tools. You also pay when your programs call on Claude’s intelligence to do some work. Each use of the search tool, for example, costs a few cents. I think it’s worth it. Claude and me, we’re proud of these!
Project 1: Smart Collections search
I gave Claude spreadsheets downloaded from the Little Compton Historical Society’s PastPerfect museum collections managment system and asked it for a smart search. Took a good bit of tweaking. How to have the search return an appropriate number of objects? How to get the images from the web? How to insure security both for the Anthropic API and for the collections data? How to properly limit use of the API? Claude added the “Curatorial Summary,” which I like. Not perfect, but pretty amazing, I think.
Search Our Collections →Project 2: Interactive map of historic sites
My second coding project was building an interactive web map of Little Compton historic sites. I started with a 1990 Rhode Island Historic Preservation Commission publication, available as a PDF. I asked Claude to turn that into a spreadsheet, and then to turn the spreadsheet into an interactive map. I threw in every feature I could imagine: color code by date, a smart search that finds sites based on their descriptions (all the stores, or all the brick houses), the ability to use the map to add images, make edits, and even add new sites, and more. It did a very nice job, I think. It came up with the categories. The one thing I couldn’t get it to do was geocode the addresses; I had to find a site on the web to do that.
Explore Historic Sites →Project 3: Audio interactive: Talk with Sarah Soule Wilbour
Sarah Soule Wilbour was a real character, one of the most interesting people of nineteenth century Little Compton. She was an abolitionist, a feminist, and a diarist. I gave Claude the diary and it to create an interactive that might be used in an exhibit. Understand the way she thought, I told Claude, and let a visitor talk to her. I gave Claude several books on the history of the town for backgound. Claude did fairly well. The technique is interesting. Claude created a “personality” system prompt for Mrs. Wilbour: “You are Sarah Sowle Wilbour (1803–1892), an elderly woman living in Little Compton, Rhode Island. You kept a diary from 1881 to 1891. You speak as yourself, in the first person, in the plain, direct, warm voice of a 19th-century New England farm woman. You are sharp-minded, opinionated, deeply caring, occasionally wry, and profoundly interested in history and the world.” A fair description. The prompt also includes not answering questions about things after 1891, or to do so only by analogy. Overall, a useful interactive.
Note: this works best in Chrome; Safari blocks microphone and voice access.
Converse with Mrs. Wlbour →Project 4: Little Compton Mini-golf!
This is a fun one. Six holes, vaguely Little Compton-history themed. Works on an iPad, too!
Play Little Compton Golf →Project 5: Download and transcribe Revolutionary War pension files
This is a practical one; make a nice download of a pension file with the handwriting transcribed. You’ll need a National Archives API key and either a Gemini or an Anthropic API key and the NAID (National Archives number) of the file. Give it that information and you’ll get back a pdf of the entire file and a Word document with a transcription and summary. As always: you’ll want to double-check the transcription and summary. In my experience, as of March 2026 the Gemini description is much more accurate. NOTE: Project 7, below, is a better, more general-purpose downloader. Many NARA files have citizen-archivist transcriptions and/or “extracted text,” that is, AI transcriptions. That version finds those, and only offers to send the file to AI transcription if one doesn’t already exist.
Transcribe a pension file →Project 6: Deed Chain Tracker
[under construction]
Track deeds through time →Project 7: General-purpose NARA downloader
Give it a National Archives API key and the NAID (National Archives number) of the file you’re interested in. It downloads the crowd-sourced transcription, if there is one; if not, the NARA “extracted text,” if there is one; if neither is available, it offers to send it to an AI tool for transcription. The files and transcriptions are available on the NARA website, but very difficult to read and search. This tool makes a very nice document of them. I put this to constant use in my work on Revolutionary War pension applications.
Download from the National Archives →Stone wall declination
Little Compton has many stone walls. Some say more than any other town. I’ve been working with the Stone Wall Stewards to document them. (More on that here.) But I’ve always been curious about how to date them. Might the angle to north tell us something about that? I asked Claude to write me a program to show the angles of walls in different areas, drawing on Elliot Vosberg’s RI Stone Wall Mapping Site. Click on the button below, select an area, and you’ll get back a histogram showing wall angles. The last part, the inferred date, clearly doesn’t work yet. But an interesting too. More work and analysis still to come.
Analyze stone wall declination →