Latest

A year after the first dotfiles post, my arch-config repository looks very different. This post traces the timeline of changes, compares the current setup against the 2025 snapshot, and explains why the AI-agent phase reshaped the whole configuration from a personal-app showcase into a minimal substrate for agent-supervised, browser-first work.

We used to split tasks by hand and launch agents one by one. Claude Code dynamic workflows (ultracode mode) now write their own orchestration script and run a fleet of subagents, shown through three real workflows that built our iread RSS reader, leaving asynchronous agent-to-agent communication as the next open problem.

After months of building tools and workflows on top of OpenCode, we are moving the daily driver back to Claude Code. The reasons are concrete: model quality after the GPT account-security change, ban risk when routing Claude through OpenCode, prefix-cache misses that burn the weekly quota, and a browser-wrapper experiment that did not return enough value to justify its complexity.

After nearly one year using Arch Linux, my workflow has shifted again: from terminal-first minimalism toward a browser-first system powered by web applications, Surfingkeys, and AI agents controlled through web interfaces.

The next step after vibe coding is not removing humans entirely, but removing them from every inner development loop. When agents can verify more of their own work, multi-agent systems stop being blocked by one human reviewer.

A practical breakdown of the two main paths for AI agent token access — pay-per-token API and subscription code plans — with real cost numbers and a concrete recommendation depending on whether you work solo or multi-agent.

md2video is a small but important experiment in agent-first software: the main input is a prompt file for the AI agent, not a GUI form or SDK call, and the result is a narrated slide video generated from markdown-based source content.