How to Amplify Your Edge with AI. Automate Your Tasks to Close the Loop.
by Jacob Koenig
3/23/26
I encoded my best practices into a system that improves itself while I sleep.
The ability to automate tasks with AI is no longer the differentiator. The tools to run agents overnight and have systems maintain themselves while you sleep are available to anyone willing to set them up.
The edge lives in what you’re amplifying: your own best thinking, encoded into a system that carries it forward and makes you sharper every cycle.
Three weeks ago I wrote about a loop I utilize in my personal AI system. This system began its life as “CloserEdge,” a negotiation coaching tool I built while running an M&A deal closing team, and has now evolved into a modular system with a persona I’ve dubbed “Eji.” It’s grown to encode my best practices so I can call on my own best thinking on the fly. Anyone can do the same:
Learn something new → encode it into AI → practice it daily → sharpen your skills → discover what else to improve → encode that too
I’ve been calling the resulting depth stereoscopic vision: two lenses, human and AI, unveil new depth when both are open. The point was never to replace your own thinking, but rather to amplify it at its best.
What’s changed since then is that the loop has gotten a lot tighter.
What does it mean to go from a project folder to a Personal Operating System?
In February, I restructured my AI coaching system from a set of 20+ markdown files duplicated across several Claude project folders into a set of modular skills. That was a meaningful half-step, but I was still reliant on the project folder memory for context recall and the persona’s interlocks to organize between skills. Updating this system was a huge chore, and so it only happened rarely. Both lenses were open, but they were out of sync.
The catalyst for the next leap was a YouTube video by Simon Scrapes called “How Smart People Are Using Claude Code Skills to Automate Anything.” It describes an architecture for building an agentic operating system designed around brand context for a business. I adapted the entire framework for personal use, replacing brand context with my own coaching philosophy, negotiation frameworks, and writing voice.
Everything moved into a modular skill architecture that I own and archive locally. Two skills form the foundation: the “Eji Core” and the “Core Context.”
The Eji skill is the soul. It carries the coaching philosophy, the persona, and understands how and when to deploy all the other skills.
Core Context gives it its memory: an archive holding knowledge of the people I’m working with, my ongoing threads, and key reference documents I need to access my second brain on the fly.
How does a personal operating system actually run?
Every other skill in the Eji system, from negotiation preparation to writing voice to communication profiling, reads from those two core skills. And it can call upon others too, even those that have nothing to do with work. For instance, the “Suno Songwriter” tool is one I use with my daughters to compose AI songs in Suno.com, complete with artist personas and genre guides.
The cross-referencing is the part that project folders could never do. Because every skill reads from the same core context, the system connects threads that I would otherwise have had to explain from scratch in every new conversation.
And now I’ve automated the update cycle. Scheduled tasks run every night via CoWork. While I sleep, the system reads my productivity data from my vibe coded habit and task tracker, connected via Supabase, and then cross-references it against the context archive. It identifies what’s changed and what’s stale, and catalogues the whole skill system, keeping a local copy current and holding my archive if I ever need to backtrack.
The daily five-minute review every morning saves hours, because Eji is always ready to take things on the fly. I can paste in a text message, say who it’s from, and the system already knows the full context: the situation, the history, how I’ve been thinking about it, and what I’m trying to achieve. It loads the relevant skill and gives me a response I can use immediately, along with coaching on how to handle it naturally next time. I don’t repeat myself, I don’t re-explain, and the system is always ready to help tease out my best thinking whenever I need it.
Why does the loop matter more than the model?
This is the philosophical core, and it connects to something John Nosta wrote recently about AI carrying the continuity of thinking itself, the idea that "thought" now continues without us. While I sleep, the system continues my thoughts by cross-referencing my context against what’s changed and preparing the next cycle. When I wake up, I continue its thoughts by reviewing what it found, confirming what’s real, and going out to practice with sharper tools than I had yesterday. We continue each other’s thinking.
Any frontier model’s output naturally guides toward the median, toward “good enough.” If you’re relying on the model alone, you’re getting the same output as everyone else. To push past that, you put your best thinking into the system. You build skills for how you negotiate, how you write, how you evaluate opportunities, how you prepare for hard conversations. Then you keep evolving those skills through practice so they get better as you get better.
The encoding isn’t a one-time setup. It’s an ongoing process, and it’s actually one of the more satisfying things I’ve done with AI, because it forces you to articulate how you actually operate and then test whether that’s as good as you think it is.
Generic AI is optimized to tell you what you want to hear, which means it can give counterproductive interpersonal advice. That's why I built Eji to challenge my assumptions, flag what it doesn't know, and hold me to my own standard. There's no replacing a trusted third party who can give direct observation that adds signal such a system can't see. But when that person isn't available, Eji is the next best thing. At the end of the day, the conversation is still mine to have.
A mirror just reflects your words back. AI works more like a prism: your input passes through a structure that refracts it, and what comes back is richer than what went in. By encoding your skills and persona, you shape the prism. By automating the loop, you put it in motion. It evolves alongside you, and the depth it produces grows with every cycle.
How does the feedback loop actually work in practice?
Compounding interest works because the returns generate their own returns, and this is the same dynamic. Every time I handle a situation well, the system encodes what I learned and surfaces it the next time the context matches. The gap between someone running this loop and someone prompting AI from scratch widens with every cycle.
Even at home, conversations with my wife that could have triggered defensiveness are things I’m now prepped and ready for. The daily context update reflects back to me where I stand across every thread in my life, including the personal ones, and helps me add awareness for what may be coming. I show up to hard conversations grounded instead of reactive. And when it works, the system picks up why.
On the professional side, last week I had a high-stakes call come together on short notice where I needed to reframe someone’s expectations and redirect toward an outcome that worked for both sides. Because the context was already loaded, the system had me prepped in minutes with the right negotiation framework and an opening move I could adapt on the fly. We got to a better result than we would have otherwise, and what I learned from the conversation fed back into the system that night. The prep for the next one will reflect what I discovered in that one.
The loop compounds
Closing the loop is what freed me from arduously maintaining the system so I could instead focus on living with it. I'm more present at home, I prep for high-stakes calls in minutes because the context is already there, and I make songs with my daughters on the same platform that coached me through a negotiation that morning. The system meets me every morning with a clearer picture than the day before, and every day I work with it, that picture gets sharper.
What works for companies and brands also works for individuals. I built this for myself, around the frameworks and decision patterns I've developed through years of practice. Simon Scrapes showed how smart people use Claude Code to automate their businesses. This is how I use CoWork to amplify how I think.
The prism keeps refracting and the depth keeps growing, because the practice never stops feeding it. And more of it happens now without me at the keyboard.
If you want to try the universal Eji package or compare notes on what you’ve been building, reach out. jkoenig@komcp.com
This article was also posted separately on LinkedIn: https://www.linkedin.com/pulse/how-amplify-your-edge-ai-jacob-koenig-nbzwe