I successfully did what I planned! I spun up a new Ubuntu Desktop VM (inside my 96GB RAM home lab) and let Claude Code control it from the beginning. I installed n8n and let Claude Code control n8n too — I literally didn't create any workflow. I wired Claude Code to Linear (via n8n), and now, at 1AM, I'm still instructing Claude Code via Linear issues and comments. Can't wait to see how far this can go, woo hoo!
It's so chaotic to communicate with AI agents via Telegram. It's only 4 agents, but the chat is already exploding. Moving to Discord — each of my slice of life will have a separated set of agents / Discord server / n8n instance. Two first ones are my 'working for employer' setup and a control plan setup, which oversees the 'working for employer' setup and more setups later!
Tried out OpenAI's GPT 5.2 with Cursor on an MCP server initiative. Surprisingly comparable to Claude Opus 4.5 so far.
Completed v0.1.0 of github.com/stanza-go/log. Implementation was unexpectedly enjoyable.
The prompt of the year...when I'm talking to my AI agent: "Please help take a deep dive into this and ask me a bunch of questions, one by one, to help me clarify, explore, and expand my ideas. Thanks!"
Resuming work on a Go package with minimal dependencies. Launching github.com/stanza-go/config as the first project.
One thing I love about AI agentic coding assistants is how they let me try and fail so much quicker and safer without frustrating me.
Company-approved GitHub Action now supports my workflows. Remaining minor issues in runtime and indexing, but we're getting there.
Anthropic acquires Bun. Can't wait for Claude Code to get even faster.
Shifting from VM-operated AI agents to GitHub Actions due to company networking restrictions. Sometimes constraints lead to better solutions.