categories · development

Development

2026

04-01 The Words I Don't Use aiagents

2025

08-07 Momentum: A Workflow That Helped Me BuildPackaging six months of workflow iteration into momentum - a Claude Code workflow that actually ships by treating AI as a compiler for operator expertise. claude-codeworkflow 07-30 The Real Cost of Claude Code SubagentsAfter burning through tokens with subagents, I learned when they're worth it and when custom commands work better. aiclaude-code 07-24 F.O.C.U.S.A meditation on constraint as capability aillm 07-18 Don't Let LLMs Run the ShowWhy LLMs excel at building software but struggle when put in control of runtime decisions and execution flow. aidevelopment 07-10 How I Made Claude Code Actually Ship Software: A Systematic Workflow That WorksBuilding a systematic Claude Code workflow with commands, quality gates, and evidence-based completion that consistently ships working software. claude-codeworkflow 07-09 Force Your LLM to SpikeHow forcing LLMs to perform research spikes before implementation can induce engineering behavior instead of just coding. engineeringllm 07-04 Building My First Claude Code Hook: Closing the Hope and Pray GapHow I discovered that the missing piece in my Claude Code + Desktop workflow wasn't better context management—it was finally having guardrails that actually know my standards. claude-codehooks 06-27 Claudex: From Ad-Hoc Context Creation to Repeatable GenerationBuilding on the three-layer memory system to eliminate the constant re-explaining of development preferences. aidevelopment 06-18 The Interview Pattern: Why AI Should Ask Before It ActsA short exploration of why letting AI interview you before implementation leads to better outcomes than discovering problems after the fact. aidevelopment 06-10 Beyond Prompting: Why LLMs Break Down on Well-Architected Code and How Composition Saves DevelopmentHow we discovered that LLM limitations reveal fundamental truths about software architecture in the AI era aidevelopment 06-09 The Architecture of Laziness: Why LLMs Are Fundamentally Designed to Cut CornersAI-assisted development tools consistently cut corners on complex work. The problem isn't training or prompt engineering—it's architectural. aidevelopment

2024

11-26 Sayable: Because AI Text Should Sound Good TooA quick experiment in making AI output more TTS-friendly aipython