LLM-First: Software for Agents as Operators
What if my software was designed for agents to use it, not just for agents to write it? Five tools, one pattern, and the factory perspective on LLM infrastructure.
Blog
Thoughts on web development, design, and technology.
The Operating Manual
How I actually use Claude Code without maxing out my tech debt. Roadmap discipline, plan mode, CLAUDE.md as living memory, and why the model is not the bottleneck anymore.
LLM Coding Agents in 2026: A Journey from Skeptic to Realist
LLMs in 2026 are excellent at writing code you already know how to write, useless for code you don't understand, and dangerous if you forget which category you're in.
Logler Web: Visual Log Investigation
A Vue 3 frontend for logler's Rust-powered log analysis. Six specialized views — log viewer, hierarchy tree, waterfall timeline, SQL editor, metrics charting, and event correlation — backed by 29 FastAPI endpoints.
Logler: 246x Faster
Three targeted optimizations cut hierarchy building from 86 seconds to 349ms. A BTreeSet prefix index, cached investigators, and capped sampling — with before/after measurements proving every claim.
Procler: Designing a CLI for AI Agents
What changes when your primary user is an LLM. Structured JSON output, self-describing commands, idempotent operations, and a config system that explains itself — design decisions from building an LLM-first process manager.