✳︎ MY AI GUIDE
Build your own AI magic diary
The open-source project that turns a reMarkable Paper Pro into Tom Riddle's diary. You write a line with the stylus, your ink fades into the page, and a vision AI writes back to you in flowing handwriting.
Built and open-sourced under an MIT licence by developer Maxime Rivest. Everything below is the plain-English version of his setup.
Get the code on GitHub →✍️You write a line on the page with the stylus, then pause for a moment.
🌫️The page drinks your ink: your handwriting slowly fades away on its own.
👁️It sends a photo of your page to a vision AI that can actually read handwriting from an image.
🪶The reply streams back onto the e-ink screen letter by letter, in an animated cursive hand, as if the page is answering you.
penCaptures your strokes and snaps each finished page.
oracleThe vision AI that reads the page and writes the reply.
quillDraws the answer back onto the e-ink display, live.
Before you start
What you need
!A reMarkable Paper Pro in developer mode. This is the one hard requirement. It is the only device the project has been tested on (the newest colour e-ink model), so other tablets are not supported.
!A vision-AI API key. It works with any OpenAI-compatible endpoint: OpenAI, OpenRouter (which also gives you Claude and dozens of other models through one key), or Groq. Prefer full privacy? A local model runs on-device too.
✓A little comfort with the terminal. You paste a couple of commands. A helper tool called remagic walks you through most of it.
✓The code itself is free under an MIT licence. You only pay for the AI usage (a few pennies per exchange) and, of course, the tablet.
The Setup
Five steps to your own diary
Tap each step as you finish it. Do them in order, top to bottom. The repo README has the exact details if you get stuck.
0/5
1. Put your reMarkable in developer mode
One-time unlock
On your reMarkable Paper Pro, turn on developer mode and install a launcher (AppLoad). The helper tool remagic can walk you through this with a single command if you would rather not do it by hand.
2. Install the diary with remagic
A verified download
With remagic set up, install the project in one line. This does a checksum-verified download straight to AppLoad. You can also grab the prebuilt bundle or build from source (both are covered in the repo).
remagic install riddle
3. Add your vision-AI key
This is the brain
Point the diary at any OpenAI-compatible vision model and paste in your API key. This writes your oracle.env. OpenAI, OpenRouter and Groq all work, or a local model if you want everything to stay on-device.
remagic config riddle
4. Open The Diary
Launch it on the tablet
In AppLoad, tap Reload, then open The Diary. That starts all three parts together: the pen that reads your strokes, the oracle that thinks, and the quill that writes back.
5. Write to it and wait
The magic bit
Write a line with your stylus and pause. Watch your ink fade into the page, then a reply appear in flowing cursive. That is the whole trick, and it is genuinely a little bit magical.
That is your own AI diary running. Now write it something you would never type into a search bar 👇
Day One
3 things to write first
Once it is running, grab your stylus and try one of these on the page.
1Ask it a real question. Write something you are genuinely turning over, and watch how it answers in its own hand.
2Start a story and let it continue. Write the first line of a tale and see the diary carry it on in the same cursive.
3Write down a worry. There is something oddly calming about a page that fades your words away and gently writes back.
💡The magic moment: the first time your own ink fades and a reply starts curling onto the page, it genuinely feels like the diary is alive. That is the whole point of the project.
Good to know
Honest answers
For this exact project, yes. It has only been tested on the reMarkable Paper Pro in developer mode, so older models or other tablets are not supported and you use it at your own risk. It is a community mod, not an official reMarkable feature.
Any OpenAI-compatible vision model. The repo shows setups for OpenAI, OpenRouter (which also lets you use Claude and many other models through one key) and Groq. You can also run a local model on the device so nothing leaves your tablet.
The code is open-source under an MIT licence, so the software itself costs nothing. You pay for the AI usage, which is usually a few pennies per exchange depending on the model, and of course you need the tablet.
You can read all of the code, and the idea is refreshingly simple: capture the handwriting as an image, send it to a vision model, then stream the reply back. If you are handy, the same loop can be rebuilt on other hardware, but the ready-to-run version is reMarkable Paper Pro only.
Each finished page is sent to whichever AI endpoint you configure, the same as any AI you paste text into. If you want it fully private, use the local model option so your pages never leave the tablet.
You’ll need a vision AI
The diary is only as clever as the model reading your page. Not sure which vision AI to use, or how to get a key? That is exactly what we help with.
Explore AI guides at myaiguide.co →