It has been a few months since my last post. <\/p>\n\n\n\n
And in that time, I seem to have gone down quite the rabbit hole…<\/p>\n\n\n\n
It started with a simple frustration
I wanted my OneNote notes outside OneNote
Which led me to Markdown
Which confirmed that open formats age better than proprietary ones
Which led me to thinking differently about knowledge systems
Which led me to connected notes and documents
Which made me wonder what else in my digital life could be connected
Which led me to thousands of bookmarks I’d been collecting for years<\/p>\n\n\n\n
Which made me realize how rarely I retrieve information from this knowledge base
Which led me to attempting AI-powered summaries
Which quickly became an AI-powered classification system
Which became an idea for a bookmark ingestion engine<\/p>\n\n\n\n
Which led me to APIs and complex prompt engineering
Which led me to structured outputs and JSON
Which led me to Python
Which made me realize coding was definitely a means, not the end<\/p>\n\n\n\n
Which led me to GitHub and how every experiment deserves a safety net
Which led me to daily debugging and frequent commits
Which reminded me of the importance of thoughtful design requirements
Which led me to scripts and workflow automation<\/p>\n\n\n\n
Which led me to my account getting suspended on GitHub!
Which changed how I approached personal software projects
Which led me to Linux, WSL and the command line
Which led me to virtual environments, package managers and dependencies<\/p>\n\n\n\n
Which led me to Git and how it works on local machines
Which led me to AI-driven coding assistants
Which led me to compare LLM models – free, paid and open source
Which taught me that reasoning, coding and planning are different AI problems<\/p>\n\n\n\n
Which revealed that building with AI has a much steeper learning curve than most folks imagine
Which drove me to think in terms of modular and reusable architecture, not one-off scripts
Which led me down context windows, token limits and specific tool calling
Which made me realize that every AI model has its own personality, strengths and blind spots<\/p>\n\n\n\n
Which taught me about data privacy tradeoffs and when using a local model is the smarter choice Surprisingly, somewhere along the way, I started having a whole lot of fun! The kind of fun where you start after breakfast, glance at the clock, and discover it’s evening. The kind where every solved problem sparks three new ideas.<\/p>\n\n\n\n I haven’t felt this immersed in learning – and this energized – in more than thirty years. The last time was in the mid 1990s, when I locked myself in my room for 16-18 hour days over many months, and taught myself computers, because I just couldn’t stop exploring that future world. <\/p>\n\n\n\n And, I have a feeling this is only the beginning…<\/p>\n\n\n\n
Which made me appreciate that the hardest part is not writing code – it’s designing systems that can evolve<\/em>
Which made me realize that AI doesn’t replace clear thinking – it actually demands it
Which reminded me that good architecture starts long before the first line of code is ever written
Which has turned all this from a project to an ongoing conversation between me and AI<\/em>…<\/p>\n\n\n\n
\n\n\n\n