For me information overload is a real problem. That's not just in general.
Beginning last year, I developed serious discipline around personal recordkeeping. That's on top of what I'd already been doing prior to that.
I have journaled regularly for over twenty years. These are in the form of grade school notebooks filled with thoughts, events, etc. I now have two dozens of these sitting in a box in my closet.
Then in 2016 I went digital. Shortly after I got out of the military, I began the tedious process of transferring my handwritten words to the digital realm. I made it fairly far (completed at least 5 or 6 years worth of journals) but I never finished.
When 2025 arrived, AI had matured to become the ultimate archive tool. I was no longer limited to just recording my own observations and experiences, I could hear them talk back to me.
What a revolution! For a lover of ideas and words, this was a dream come true. But it came with a problem: an overwhelming volume of informational ore.
No longer did I only have my own writing as evidence of my thinking and activity. Now there were novel-length AI exchanges, photos, podcast transcripts, code, and video transcripts (my own and others) that entered the information flow. I quickly began to drown under the deluge of data.
So things got lost. A brilliant idea for a story, app, or service on Tuesday, was washed away like loose beach sand by Thursday. A clear character profile based on a spontaneous interaction was practically erased. And worse yet, I wasn't doing anything with everything I was accumulating. I had become a hard drive hoarder, suffocating under mountains of information of my own making. There had to be a better way.
Building the Pipeline
The chef analogy arrived late in the day, after the pipeline was already running. I was thinking out loud, working to explain what I had just done, and the metaphor for organizing and synthesizing my prima materia with the help of AI surfaced on its own: it's like having a robot prep cook. You walk into the kitchen and all your ingredients are out, your mise en place is done, and all you have to do is fire up the stove and get busy.
The idea for the system was simple:
- have AI read my previous day’s notes,
- summarize, synthesize, and index the material into a draft useful output
- send me an email with the results
I'd tested its capacity to perform all of these tasks individually over the past several years, so I knew all I had to do was assemble it.
That's why I spent yesterday morning building this.
The result was a Google Apps Script that:
- reads the most recent journal in a Drive folder,
- sends it to Claude with a detailed prompt,
- produces a Morning Novelist Brief
- Sends me an email with the link to the Google Doc
This process runs daily at 4AM and is ready for me by the time I'm up and ready to work. It costs about twenty-five cents per run.
I'd built my own personal novelist assistant.
Why it took me so long to build it
I have been wanting to build something like this for years. Not exactly this. But something— a system that solved the information clog that I experience daily. These days I generate notes faster than I can read them. I record voice memos and then forget what's in them. I write a journal entry and a week later couldn't tell you what was in it without searching like a man who's lost his keys before work. The mass of my own work has been outpacing my ability to use it for a long time.
So why did I wait so long to create this system?
The surface answer is that the AI technology wasn't capable enough. Tiny context windows, generic output were no good for a sustained workflow.
But the real answer is that I was intimidated by the money.
I always worried the bill for using the metered AI would be too high and what I was getting wouldn't justify the expense.
But turns out in my situation, finances were a flimsy excuse.
The whole project, including the test runs across multiple debugging cycles, cost about a dollar. This morning the bill was about thirty cents. At the projected pace — five briefs a week — I'm looking at six or seven dollars a month. For less than I spend on two trips to Dunkin’ Donuts, I now have a personal assistant trained on my own thinking that briefs me every morning.
The real obstacle was mental (isn't it always?)
So the money was never the real obstacle.
The real obstacle was that I had categorized myself early in life as a person who used tools but didn't build them. In high school and college I was the “fuzzy” kid who, despite having skill in high level math and related subjects, and participating in my high school’s first robotics team, told himself understanding tech was out of reach.
Writers don't build automations.
Novelists don't write code.
There's a clean role I'd been performing, and the role didn't include this kind of work. I'd written half a dozen books with AI assistance, but always as the author commissioning the work, never feeling like a true engineer, despite having experience with coding pre-AI. The line was invisible to me until I crossed it.
And what happened upon crossing it?
Nothing dramatic. I sat at my desk with three AI assistants open in tabs. ChatGPT scaffolded the architecture. Gemini wrote the first pass of the code because it has cleaner Google Workspace integration. Claude caught Gemini's outdated model string and rewrote the payload. They all became co-engineers. The whole process was conversational. None of it required me to be a paid or even experienced developer. It required me to be willing to be a beginner for six hours.
Writing my own story about AI
Online, AI (like politics) is a narrative of extremes: is AI overhyped or going to the moon? is the bubble popping or are we headed for eutopia? are these companies finished or are they ushering in a new are of wealth like the internet before it?
The boom-bust cycle narrative is everywhere and has pernicious gravity. It can easily draw you in and crush you.
What many don't understand is that the majority of the discourse is aimed at investors and ideologues. It has almost nothing to say about the quiet category of people who are using these tools to build small machines for their own specific problems. Because we're too busy to post about it. The work is too undramatic to engage. A novelist in Las Vegas built himself a synthesis layer between his journals and his publications. That's not a good story or headline.
A metaphor that came to me later was Home Alone. Early in the movie, Kevin McCallister stands in his empty house and says, "I made my family disappear." The line is panicked as the weight of the situation sinks in. However, later he says it again— same words, different inflection— and now it means I can do whatever I want.
That's the position I'm in with AI.
The first version of the line was the version I'd been living for years: intimidated, uncertain, watching others build, in an exclusion zone of my own design. The second version is what happened yesterday: an open road of opportunity with very little traffic. It's that way because almost nobody else is looking. Most are paying attention to the online shouting matches outside the window instead of looking ahead. The Silicon Valley evangelists are loud, the doomers are louder, and in the middle there's a silent, uncrowded territory where individual practitioners are building individual tools for individual problems, and almost no one is reporting on it because none of us have time.
The brief is running now. Tomorrow morning at 4AM it will read this entry, identify what I've been circling, and propose what to do with it. The prep cook has its first shift before dawn, ready to transform mental clutter into a treasured meal.
And when it's finished I’ll just wake up and cook.
