I am giving this blog another shot. Over the past few years, I've slowly adopted more and more AI into my workflow as the technology evolved.
That shift led me to a stark realization: the messy trial-and-error, the dead ends, and the human decisions that actually make software development valuable are now getting trapped in hidden, private chat windows.
Surfacing those specific insights is exactly what I want to share with the world. But before I start publishing again, I want to talk about my philosophy on writing technical content in the AI era.
The internet is already overflowing with AI slop. We've all seen those painfully boring, generic posts. If you just ask an LLM to "write a blog post about X," you get garbage.
Why? Because the AI lacks the one thing that makes technical writing actually worth reading: Context.
The Problem with Throwaway Chats
For years, platforms like Stack Overflow were our public squares for knowledge sharing.
Today, generative AI is just easier to use for solving problems, but it's created a huge blind spot. We aren't posting our weird edge cases or hard-won debugging tricks in public anymore.
When a feature ships, you close the AI tab. Technically, the raw history of how and why you solved the problem still lives somewhere in your chat logs.
But realistically, no one goes back to dig through thousands of lines of AI dialogue, and a raw transcript isn't very useful on its own anyway.
If you try to rely on the AI to just summarize the feature for you later, it will only generate a bland, high-level overview. It completely misses the real pain points because your synthesized context was left behind.
Using AI to skip the tedious parts of writing is great, but generating words without that hard-won substance just adds to the noise online.
Documenting the Journey
Using an LLM to automate the writing process is totally fine, BUT the topics and content of the blog must focus on real substance derived from actual context. Raw context isn't actually that useful on its own if we don't actively document the journey where it matters - surfacing the actual solutions, the insights, and the human decisions.
To write something meaningful, I need to capture those decisions before they disappear. I take a lot of inspiration here from Nicole van der Hoeven's ideas on Learning in public and Working with the Garage Door Up.
Adapting those concepts to the AI age, I like to think of my approach as "Working Out Loud."
I actively document my journey as I work. The real substance comes from well-written User Stories, incident reports, project-specific notes in a repo, Architectural Decision Records (ADRs), or even pulling specific discussions straight out of an AI coding session. I'm even exploring tools like Hermes in conjunction with Honcho to further inject my own preferences and findings.
The Future of amoenus.dev
That brings me to the whole point of this post. Consider this a checkpoint and a restart for the blog.
Moving forward, I will be using AI to help me write. But the foundation of every post will always be the problem-solving records I actively maintain, careful curation and topic choice.
