My AI That Remembers Me
11 March 2026 · 2 min read
How I turned a generic AI into a contextual co-brain that saves me 15 minutes every morning
The Morning Problem
Every morning, I did the same routine: Open Gmail (personal + work), check Google Calendar, scroll through Notion tasks, mentally plan my day.
Takes about 15–20 minutes. Not huge, but it adds up. That's 5+ hours per month spent on “admin stuff” before I even start actual work.
Meet Vansken
Vansken is my AI co-pilot. He lives in my Telegram. I chat with him like I would with a teammate — casual, direct, no fluff.
He didn't learn this from a single prompt. He learned it over weeks of conversation, stored in a memory system that persists across sessions.
The Build: How I Set It Up
I didn't code Vansken from scratch. I found OpenClaw — an open-source AI agent framework that runs locally. Think of it as a “build your own AI assistant” toolkit.
The setup was surprisingly non-technical: Memory System, Skills, and Personality — defined his tone: casual Indonesian (lu/gua), short sentences, direct.
The Features That Actually Stuck
Morning Brief saves ~15 mins/day: Every morning at 7 AM, Vansken sends me a summary — unread emails, calendar events, priority tasks, and a daily spark.
Body Composition Tracker: I send him InBody scan numbers. He logs it, tracks trends, and gives me actual insights.
What I Learned
Memory > Intelligence — A moderately smart AI with perfect memory beats a genius AI with zero context. Every time.
Start small, add incrementally — Don't try to build everything at once. My first version just did email summaries. The rest came later.