In 2001, Wikipedia was launched, and its reception was… mixed, to say the least. Many dismissed it as unreliable compared to authoritative sources like the Encyclopaedia Britannica. Yet Wikipedia had something that Britannica never could: a scalable, living knowledge system, constantly updated and freely accessible. Just over a decade later, Britannica stopped printing books in 2012.
Today, Wikipedia content is increasingly consumed through AI answers and search snippets, rather than by visiting the site itself. The information layer is shifting again, and this change is not limited to encyclopedias. It will affect how we build, document, and support software.
I use documentation as a sample in this post because it’s where our AI-first efforts with rsyslog began—but the full vision is broader. We are working toward AI-assisted development, integrated testing, and smarter support tools—all connected by a unified AI-first pipeline. Documentation is just the first step in demonstrating what this future will look like.
I can see this transformation coming because I have lived through and helped drive many such transitions. I started in the 1980s, working with datacenters, networking, and early PCs. Later, I contributed to open standards as an RFC author and was part of the wave of technologies that made Wikipedia and similar platforms possible. I’ve seen these disruptions up close, and I’ve learned how critical it is to prepare early.
AI-First: The Broader Goal for rsyslog
Our current work on rsyslog documentation is not just about making the docs better -though that is badly needed. It is part of a broader AI-first strategy for rsyslog. In the future, people will not read documentation in the traditional way. They will query an AI assistant and get precise, context-aware answers.
But we will not stop at documentation. The long-term goal is to consider the entire lifecycle of rsyslog—from idea to design, development, testing, and support—and see where AI can assist, automate, or amplify. Documentation is simply the first building block of a much larger transformation.
Strategic and Tactical Work
For nearly two years, I’ve been investing in AI prompt engineering and testing model capabilities, not as an experiment but as strategic groundwork for this AI-first approach. At the same time, improving the rsyslog docs is tactical—we need better documentation immediately, both for users and to feed into our custom AI tools.
This work already goes beyond documentation. AI helps us identify code patterns, generate test cases, analyze configuration examples, and even suggest improvements to development workflows. It’s far more effort than writing docs the old way, but that’s deliberate: we’re mapping out the limits of current AI and building expertise to integrate it across the full rsyslog lifecycle.
Every AI-assisted change—whether to code or docs—is manually reviewed and confirmed. The results are not perfect yet, but they’re improving steadily, and we’re transparent about the process. We deliberately take the hard path now to build something much better later.
Documentation as an Interim Format
I believe that documentation as we know it is temporary. It’s becoming an interim layer—a structured knowledge base, maintained mainly so AI can access it and provide correct answers. With self-documenting, metadata-rich code, we might one day reduce or skip this layer entirely. In the future, the code will be the documentation, with AI serving as the delivery interface.
The rsyslog Assistant – A First Step
This vision is already taking shape with our rsyslog Assistant, a custom RAG-based GPT. With the improved documentation—modest as the progress may be—and carefully engineered prompts, this assistant is already better than a plain ChatGPT for support questions. It’s far from perfect, but it often points users in the right direction and delivers accurate answers.
The rsyslog Assistant is not just a support tool—it’s a prototype for the future of AI-first rsyslog development. It demonstrates how structured knowledge and AI can combine to make the entire project lifecycle—support, coding, and documentation—more efficient and approachable.
The Importance of Correct Data
AI is an amplifier. If it is trained on bad or outdated data, the output will be misleading or wrong. The internet is full of incorrect rsyslog information, which is why we are building a curated, authoritative knowledge pipeline. Our work on documentation directly feeds this pipeline, ensuring that AI tools can deliver trustworthy, accurate answers.
AI is Not for the Lazy
This approach is hard work, but it’s the only way forward. We’re not just patching the documentation—we’re building a next-generation knowledge system that combines code, structured documentation, and AI. This is what AI-first truly means for rsyslog.
I’ve seen massive changes before—Wikipedia replacing Britannica, the rise of PCs and networking, the web, mobile, open source, and countless others. AI is simply the next step. Documentation as we know it today will not survive the next decade in its current form. The real question is: are we ready to build the next layer?