Why I Am Building My Own Newsletter Tool, in Public
I pay a newsletter service for my sticker shop, KANNSTEMACHEN. A few thousand people are on the list, and I write to them rarely and at no fixed rhythm. Like almost every service of its kind, it charges by the number of subscribers, every month, whether I send or not.
Long before I started the shop thirteen years ago, I used newsletter tools and helped other people choose theirs. The questions then and now were always the same: which features, what it costs when the list grows, where the readers’ data ends up. No tool ever answered all of them, so the answer was always the best compromise on offer. That is not special to newsletters. It is how buying SaaS works.
In the past two years, the question has started to change. Especially for people who have a deeper understanding of software, it is no longer only which tool fits. It is whether they can build the one that does.
So last Monday I started building my own newsletter tool. I call it Roadletter Lite, and I want to build it in public, one stage per week. That is the plan.
An idea that always looked too big
A newsletter of my own, on my own terms, has been on my mind for a long time. But it always looked too complex, and there was never time.
From the writer’s side, a newsletter is simple. You write, you press send. But under that one button are the protocols of email, the infrastructure to send hundreds or thousands of messages, and the work of making sure they land in an inbox and not in spam. For a long time, that was not a project for one person.
What tipped it was an episode of The Every Podcast (formerly AI & I) with the writer Craig Mod. He replaced a newsletter service that cost him thousands of dollars a year with a small tool of his own, built with the help of AI on Amazon SES. Sending now costs him about $150 a year, and the tool does only what he really needs.
What has changed is the effort. With AI and a good understanding of software, trying something like this takes a fraction of what it used to. So the project asks one question: how simple can a newsletter tool be that focuses on writing, respects its readers, costs little and stays on the right side of the GDPR?
A small tool that lives on your own server
Roadletter Lite will be a small newsletter tool for anyone who writes to a list of readers: writers, freelancers, small shops, startups. You host it yourself.
Readers will sign up with double opt-in, and unsubscribing will take one click.
Every issue also lands on the web, in an archive with an RSS feed, unless you keep it for the inbox only. You write once, and it reaches readers both ways, in their inbox or on the web.
There will be no open or click tracking, not even as an option. I do not like being followed around the web all day, and I will not do it to my readers. EU and German law are also clear about what you may and may not do with a reader’s data. But my main reason is trust: someone who gives you their email address trusts you with it, and that should not depend on a setting.
Export, erasure, retention periods and proof of consent will be built in.
Running Roadletter Lite should cost about €9 a month for 1,000 subscribers: about €7 for the server with backups, under €1 for sending four issues, about €1 for the domain. Even with 25,000 subscribers and four issues a month, it should stay under €25.
The code is source-available under the O’Saasy License. You can read it, run it on your own server and change it.
What it will not be
There will be no hosted version. It runs on your server, or it does not run.
There will be no installer and no setup wizard either. You need to be able to follow a technical guide, or know someone who can.
There will be no automations, no segments, no A/B tests and no growth tools either. It stays that small on purpose: one list of readers, one issue for all of them, and a website that keeps the archive.
Eleven weeks, with the work in plain sight
There are eleven stages, numbered 0 to 10, one per week, from last Monday into the first days of January. At the end of each stage there will be a post on this blog and a release on GitHub.
It starts with the foundation and ends with version 1.0, with my own newsletter moving onto it. The roadmap shows every stage.
I am building it with Claude Code, and I write hardly any of the code myself. The design and the interface copy come from AI too. Claude and I plan each stage together in writing, then Claude writes the code, and I review, correct and decide. My part is the product: what goes in, what stays out, and whether it is good enough. I am responsible for every line, including the ones I did not type.
It all runs inside an AI product development framework of my own, built and refined over some time. Every stage tests the same question: what can the AI do on its own, where does it need a human sparring partner, and where is an experienced human simply better? I will write about that framework as the project goes on.
Roadletter Lite is a side project, not an attempt to get a whole app out of a single prompt. The AI will make mistakes, and the posts will show where I had to correct it.
It is an experiment, and I do not know yet whether eleven weeks are too few or too many. We will find out.
Week one, without a line of code
Stage 0 had no application code at all, on purpose. Before the first line, I wanted the scope and the decisions written down where anyone can read them.
The repository is public now. It holds a README with what Lite will and will not do, the license and the rules for contributing: bug reports and questions are always welcome, but the bar for new features is high. Roadletter Lite is supposed to stay small and simple.
The repository also holds the first five decisions, each on one page, from the tech stack and the license to no tracking and an interface in English and German from day one. Each one says what it makes easier and what it makes harder.
More languages will follow. If you would like to add yours, get in touch.
Last week ended with a first release on GitHub, v0.0-start. There is nothing to install yet. There is only a plan for now.
Where to watch it happen
Everything lives on roadletter.com: the roadmap, the posts as they come out, and every place to follow along.
Stage 1 starts today: the foundation of the app, with sign-in, English and German, and light and dark mode. The next post comes out when that stage is done, at the end of the week.
If you write a newsletter, or build tools for people who do, follow along. And tell me: what would a tool like this have to do before you trusted it with your readers?