Blog
Articles, insights, and deep dives from the Arcanada team.
Why I Started Building Arcanada
Food & Ammunition, Initiative 10, and the everyday responsibility of work and family all called for the same things: research, distinct roles, memory, and a way to keep going.
One Person and a New Scale of Software Development
AI agents let one person carry a project farther from idea to working system. For Pavel, that shift led to a question: what should he spend the next twenty years building?
A Chrome Extension That Saves the Entire ChatGPT Conversation
I often talk to ChatGPT while training, then pass the complete conversation to Codex. That workflow needed a small open-source extension.
Arcanada month three. 823 core tasks and a Mac I can turn off
Arcanada report for June 15 to July 18, 2026. Month three delivered 823 core tasks and 3,711 routine tasks, costs reached $807, and total investment was at least $2,240. Development moved entirely to servers.
Cubrim-2: The Global Addresser and a Shared Language for Data
Cubrim-1 is already a measured archiver. Cubrim-2 asks a separate question: can devices share a verified memory in advance and transmit only what is not there yet?
Building the Binary Is Only the Beginning
Cubrim’s first public release exposed the work between a finished binary and software a user can actually download, install, and run.
Cubrim. Or How Dreams Come True with AI
The story of how an old student-era idea about data compression sat in my head for years and finally came alive — with the help of AI agents. Today Cubrim reached the top of the world compression benchmarks.
After Self-Healing: Why Agents Need Autonomy Levels L6 and L7
In the previous article I mapped the telecom L0–L5 autonomy scale onto AI agents. But for agent systems, L5 is the ceiling only for a single agent. Beyond it comes self-assembly: an agent that builds other agents (L6), and an ecosystem that develops itself within human-set bounds (L7). I work through both levels, separate them from an ordinary multi-agent workflow and from AGI, and show where Arcanada fits.
Show Me How You Will Do It?
The engineering war against agent hallucinations — not the abstract kind. The simplest question I ask an agent before every step, and why it changes how the model works.
The Human at the Console: Procrastination, Switching, and the Long Game
The finale. Not about technology — about the human who built the system and still sits at the console eighteen hours a day. About procrastination, about how to keep from burning out by switching between projects, and about what it is all even for.