Field notes, written as I ship. Newest first.
96 posts
I shipped the same feature twice, once in Kiro and once in Claude Code. The code converged; the workflow did not. How I route between spec-driven and terminal-native in 2026.
I ran Amp and Claude Code on the same full-text-search task for two weeks. The code converged; the defaults did not. Amp acts without asking, Claude Code asks first. Here is how I route between them in 2026.
I gave one nasty order-dependent flaky test to Claude Code, Cursor, Codex CLI, Cline, and Gemini CLI. Here is which AI coding agent actually debugs in 2026, and when to reach for each.
Claude Code plugins bundle skills, MCP servers, and language servers into one install. A 2026 field log on which plugins I kept, which I deleted, and the context cost nobody prices.
I ran one real refactor through Claude Code, Cursor, Aider, and Codex CLI, ranked by safety net and review surface, not raw model smarts. A 2026 field log.
windsurf.com now redirects to Devin Desktop. I built the same optimistic-UI feature in Windsurf and Claude Code in 2026. Here is which surface I kept, and why.
I ran the same webhook-handler feature through Gemini CLI and Claude Code on one repo. Gemini CLI wins for large-repo exploration and giant context; Claude Code wins for consistent multi-file changes. A 2026 field log with my routing rule, corrected after Google withdrew Gemini CLI's free individual tier.
I ran the same pagination change through OpenAI Codex and Claude Code on one repo. Codex wins when I can delegate a boxed task; Claude Code wins when I want to steer a multi-file change live. A 2026 field log with my routing rule.
OpenCode is open source and model-agnostic; Claude Code is managed and tuned around one vendor. I ran the same invite feature through both on the same repo in 2026. Here is which I kept, and the routing rule I use now.
Quick answer (2026): I ran the same feature through GitHub Spec Kit, Kiro, and a plain plan.md in Claude Code to see which spec-driven setup actually earns its place. Spec Kit wins for writing shared project rules, Kiro wins for killing spec-and-code drift, and a hand-written plan.md wins on setup cost. The plain plan.md is the one I kept.
A three-month field log on the best MCP servers in 2026: the five that earned a permanent slot, the ones I uninstalled, and the context and trust costs no listicle mentions.
I hit Claude Code's weekly limit mid-refactor, then ran the same repo through seven alternatives for two weeks. The three I kept, and the honest catch nobody markets.
