Apps, apps, and more apps
They have a shiny design, they have attractive colours, they make you feel like a better person, the CEO of a great startup. That’s what apps do to me (and, I think, to many people). You need them to organise your life and achieve your goals. You need them for every single aspect of your life: from tracking your sleep to brushing your teeth.
Going for simplicity
I am at a point in my life where I do not want to be stressed by constant (and often useless) notifications. I do not want to be distracted by my watch asking me if I am riding a bike. I probably already know that I am.
One of the key aspects I want to improve is daily organisation. I have three kids who need to be picked up from school or kindergarten, so every movement must be well organised. I discovered online the “one big text file” pattern:
I tried it for a couple of weeks, improving and adapting the structure based on what I saw working well and what could be improved. I put almost everything in the file. Here is an example:
# 2026-01-26
**07:00** wake up and prepare for the new week.
**09:30** sprint planning meeting.
**11:30** sync for the new feature.
**12:00** 10k slow run.
**13:00** lunch.
**13:30** start working on the new feature. [TODO]
**17:30** wrap up the day.
**21:00** watching the big match.The format is quite simple: a big markdown file with an easy structure. It worked quite well. I usually write the next day’s plan the day before, so I already know what I have to do and where the friction points might be.
I was happy with my file, but maybe…
The big file is really handy. I can search for any activity I have done in the past and immediately get a result. There is no need to jump between different apps for different kinds of information. Everything is in one place: appointments, notes, todos.
Postponing an activity is just moving a line. Adding notes means adding a line and starting to write.
What's difficult with the large file is to extract meaningful information: you have to go through many lines looking for specific words and then summarise what you found. But since a large part of my workflow already revolves around LLMs (I am joking… mostly), why not add a bit of magic? That is how Ambrogio was born.
Ambrogio takes its name from the Ferrero Rocher commercials: the discreet, always-present butler who anticipates needs and quietly solves problems. It is not about replacing you, but about helping you.
Ambrogio is powered by an LLM. LLMs are very good at working with text files. You just need to teach them the structure, and then you can start asking questions about what you have to do tomorrow, which tasks you have not finished yet, and so on.
Let’s go technical
The technical choices behind Ambrogio follow the same principle as the organiser file: keep things simple, local (if you can use a local LLM), and under control. The implementation reflects that more than any architectural ambition. I wrote the app in Rust. I started using it a couple of months ago and I am still not proficient with it, so I rely heavily on AI (Claude Code) to help me write the code. You can see Claude involved in almost every commit.
I created a CLAUDE.md file that I reuse across
many projects. After several iterations, I found a way to
guide the LLM towards the structure I have in mind, and it
works quite well.
I run the app from the terminal. It is configured to use my daily organiser file. When I start it, I can ask different kinds of questions. In these early days, I mostly use it to summarise notes from the past few weeks or to prepare for specific meetings. So far, I am quite happy with how it works.
Ambrogio works with an LLM of your choice. It sends the content of your big text file to the model and lets you ask questions. That’s it.
Little tools with a bright future
I started this post complaining about apps. Ambrogio is not an alternative app, but a way out of that pattern: no dashboards, no notifications, just a tool that answers when I ask. My initial idea was to start from small sources of friction and try to solve them in a couple of hours. Thanks to the power of AI, and the experience I have built using it, I am able to produce something reasonably useful with a few well-crafted prompts.
There is plenty of room for improvement: structured outputs for todos, colours, better formatting. Kaizen is the way. Small improvements every time I notice something that can be done better.
Thank you for taking the time to read about Ambrogio. If you are interested in the project or have ideas for improvement, feel free to reach out or contribute to the GitHub repository.