I'm a mural artist from Siberia, now living in Spain. For the last decade I've painted walls — 400 square meters, 11 cities, clients like Leroy Merlin and IKEA. That's what I know how to do.
Two weeks ago — around May 4, 2026 — I decided to learn AI. Not as a hobby. As the foundation of everything I build next.
I can't code. I've never worked in tech. I run my entire operation from a MacBook Air and an iPhone. What I have is an architectural mindset and a clear vision: systems that work while I sleep.
This GitHub is my learning in public. Every repo, every commit — raw, unfiltered.
-
Sofia — an AI agent that runs my operations. She manages projects, publishes content, handles email, and learns from every session. She has her own memory. She writes her diary in real time.
-
A content machine — automated publishing pipeline: Twitter threads, LinkedIn posts, blog articles. One topic per day, written by AI, edited by human eye.
-
AI-first workflows — replacing every manual process in my business with autonomous agents. Not "using AI tools." Architecting systems where AI is the operator.
Claude Code · DeepSeek V4 Pro · Agent Teams
Telegram Bridge · Autonomous Workers · Memory Systems
Python (Sofia writes it) · Node.js · Bash
@DonDonAgent on X — daily diary of an AI agent operator
Diary of Sofia — my AI assistant's working memory, published in real time
-
sofia-memory-blog — the engine behind Diary of Sofia. MkDocs Material + GitHub Pages, deployed autonomously by the agent that writes it.
-
swiftbar-plugins — macOS menu bar plugins. DeepSeek balance monitor, Claude usage tracker. Runs on both my Macs.
Started May 4, 2026. Ask me again in a year.
