How Tech Acceleration Forced a Rethink of My Personal Stack
What started as a simple cost-cutting move turned into a re-architecture of my personal systems. Rebuilt around markdown, new tools, AI coordination, and guided by systems thinking.
I wasn't trying to build a "Personal OS." I just didn’t want to keep paying for a website nobody visited.
But one cost optimization decision led to another, and soon I was running my digital workflow differently.. Not because I had some grand vision, but because I started thinking like a systems person about my own infrastructure.
The $400 Problem That Started Everything
Wix wanted $350 to renew my static three-page website. Google was charging me $30/month for email and document storage I barely used. I was paying nearly $400/year for pretty basic digital services that weren’t adding a ton of value.
I'd been playing with new tools, mostly using ChatGPT in a browser window. Copy, paste, prompt, copy back. Then I had this tee time booking problem. Spots get released randomly and checking manually is a pain. I asked ChatGPT: "Could we build a bot that scrapes tee times and alerts me when spots open?"
A few hours later, I had working code that solved my actual problem. Now that felt valuable.
That’s when the questions (and research) really kicked in. How do I apply this everywhere?
Rethinking the Stack
I was paying premium prices for low value utilities. What am I actually getting value from and lets shift the spending toward that.
Email, websites, basic document storage, these are commoditized now. Almost free if you’re willing to manage the systems yourself. AI assistance felt like where the real leverage was.
So I started reworking my setup around that idea:
Website: Moved from Wix to Eleventy + Tailwind running on the Pi. Cost went from $350/year to $12/year for the domain. Same functionality for my needs, way less cost.
Documents: I canceled Google Workspace and moved everything to markdown files on the Pi. Easier to navigate, simple to backup, and AI tools handle structured text than Google Docs formatting.
Development: Started using Cursor IDE manage everything (markdown files, code, notes) all in one place. Reduced the need to to jump between apps.
AI Integration: I switched to Claude Desktop with MCP connections so it could read my files directly. No more copy-paste into a chat window, just tell it what to reference.
Each change brought a little more of my work into a centralized place where I could actually leverage AI. Almost like I was solving for “how do the tools work?” instead of “how do I want to work?”
What To Remember
I've always used ChatGPT's memory. Turned it on, started feeding it background on projects and preferences. Worked great initially.
But memory sprawl is real. You can see what it “stores” in the settings, but it seems to remember more: conversations, tangents, half-formed ideas. You tell it something once and it becomes part of your permanent record, shaping every future interaction.
Eventually I had to download what it remembered, edit out the irrelevant parts, wipe the memory, and reload an updated version. Even then, it still seemed to carry baggage from other chats.
I’m not sure if the answer is better memory management inside ChatGPT, or a more structured system where I control what context gets loaded, like what Claude Desktop lets you do. But that’s the direction I’m exploring.
What Changed Beyond the Tools
This isn't just a Google Docs swap. It was infrastructure thinking, asking what my systems were actually optimized for and whether that still makes sense.
Most people's digital workflows were designed for the pre-AI world. Scattered across 30 different apps, each one optimized for its own domain. Email for messages, LinkedIn for networking, Google Docs for collaboration. None of them coordinate by default.
But AI needs clean inputs and clear context. When your information is structured and accessible, AI becomes a real-time collaborator instead of a copy-paste assistant.
What This Looks Like Now
I'm in Cursor: folders for my website, tee time bot, content files, and other projects. Everything in one interface that AI can understand and navigate.
When I need Claude to help with something, I can point it to specific files and it reads them for context before responding. Last week an ex-coworker asked for a quick bio to include in an intro. Claude read my background files and wrote it. No copy-paste.
The workflow feels fundamentally different. I've had to create those context files and keep them organized, but now that the legwork is done, it's rinse and repeat.
My screen used to be two Chrome browsers with 25 tabs each, plus more Finder windows than I could count. Now it’s Claude Desktop on the left, Cursor on the right, and one Chrome window with maybe 20 tabs (email, calendar, ChatGPT, websites, etc.). Not a massive change but it's progress.
The Bigger Shift
What started as a cost-cutting move turned into a bigger project: figuring out how to set up my systems to make the best use of AI.
Getting good results still takes prompt and context engineering but I want to build a system where that happens by default. Where creating, using, and storing prompts doesn’t create extra work.
The commodity services (email, websites, storage) are just utilities. The value is in orchestration. In the intelligence that puts the data to use. That’s where I’ve tried to focus my time and spend.
What's Next
The experiments are getting more interesting now that the foundation is different. Voice input feels way faster than typing, no need to be articulate, so I’m trying to build that into the workflow.
Multi-platform search: LinkedIn, X, Reddit, news. Can I aggregate all of it into one interface?
Personal CRM: pull together my contacts, messages, and notes so I can context search, communicate, and network more effectively.
What I really want is research, learning, and communication at scale. Nothing revolutionary. Just a way to do what I’m already doing, faster and better. Apply systems thinking to personal infrastructure until AI coordination works for me, instead of the other way around.
The chat window was just the beginning. The real destination is a system where everything you're already doing gets leverage from AI.
Why This Might Matter
I know I’m not the only one trying to get better. But most of what I’ve seen is tactical: “write better prompts,” or “ask better questions and learn from the answers.”
The more strategic question is infrastructure design. How do you structure information flow so AI gets good inputs? How do you build workflows that actually scale? How do you shift your spend toward the tools adding the most value?
These are the same questions I’m working through in my personal systems where I can move fast, experiment freely, and see how it all connects.
No master plan here. Just a systems thinker who likes to tinker, working through coordination problems that feel like they’ll matter more as AI gets better.
Building similar experiments? Always open to swapping notes on systems, tools, and process!