Hi, I'm @geediab.

Running Perplexity's read-only endpoint scanner: zero catalogue findings, sixty-three MCP servers I never installed, and one freshness pattern worth knowing.

What I learned building a markdown second brain with Tolaria, git, and an AI collaborator that helps with capture and structure, not writing the notes for me.

These two plugins made Claude Code easier to use for real engineering work by tightening the loop around context, planning, execution, and review.

What happens when an engineering team goes through an AI-driven identity crisis: the grief, the fracture, and what reorientation looks like on the other side.

I let GasTown build a Toggl TUI. The speed impressed me, but the real insight was watching a Mayor agent shift from solo coder to multi-agent coordinator.

AI doesn't reduce your cognitive load. It concentrates it. That distinction matters more than you think.

A two-run case study of Entire CLI attribution: what the first test left ambiguous, what the prompt-only rerun clarified, and what still needs interpretation.

How I built a local-first memory layer that captures sources quickly and returns compact, source-backed context for AI agents.

A pragmatic guide for engineers and leaders: scale AI coding autonomy by investing in verification, quality gates, and evidence pipelines.

AI got great at coding first because it compounds. The real shift is for anyone who works on a screen: turn intent into output faster and iterate more.