Writing · Stockholm · 2026

Writing.

Notes from running an agent fleet as coworkers. Multi-agent systems, fleet operation, infrastructure underneath. New pieces shipping as I finish them.

No 01
Setup
Claude Code in tmux background sessions: a working setup

Every long-running agent lives in a detached tmux session named after its job. Six months of fleet operation distilled to one rule.

No 02
Operations
When to restart a long-running Claude Code session

Token saturation is the only signal that matters. Wall-clock time, idle time, vibe, all decoys.

No 03
Architecture
A context vault for AI agents: persistent knowledge that compounds

A SQLite-backed memory layer agents can write to and retrieve from across sessions, providers, and machines.

No 04
Architecture
Fractal context: how long-running agents stay coherent

The recursive structure that lets a single architectural rule survive across sessions, providers, and tier boundaries.

No 05
Essay
Build a personal AI operating system with Claude Code

From single-prompt user to fleet operator: the eight components of a personal AI OS, and what shipping each one taught me.

No 06
Stack
Vanilla HTML beats your framework

For solo developers and agent-driven sites, vanilla HTML beats React, Next, and Astro. The framework tax is hidden until you remove it.

No 07
Stack
Rust as the boring choice for solo tooling

Three of the four CLIs and servers I run every day are now Rust. The fourth being JavaScript is my fault, not its fault.

No 08
Stack
Marker pairs over template engines

A 200-line Rust binary that does literal substitution between HTML comments. Smaller than every templating system I've used.

No 09
Tools
The HTML file IS the CMS

Most CMSs treat HTML as a build output. Freedom CMS edits the HTML file directly via contenteditable, autocommits to git. No database, no schema, no build step.

No 10
Tools
Building an AI memory layer with context-vault

A memory layer for AI sessions: MCP-backed local SQLite, semantic retrieval, lifecycle tiers, and conflict detection. How and why I built context-vault.

No 11
Architecture
Building a self-improving agent OS

A folder tree as agent configuration, three orchestration layers, and a self-improvement loop where agents log friction that becomes the next round of issues.

No 12
Operations
Eleven PRs while I slept

The first overnight run of an autonomous agent orchestrator: 686 cycles, 11 merged PRs, 8 self-healed stalls, and 94% idle time. What worked, what didn't, what's next.

No 13
Essay
omni: an agent orchestrator for my GitHub workflow

A CLI orchestrator that turns AI agents into long-running workers against a GitHub backlog: a 3-layer master/workspace/project model with a vault-backed feedback loop.

No 14
Setup
Using Git with AI agents

A practical Git workflow for vibe coding with AI: clean branches, atomic commits via git add -p, intent-driven messages, and PR review as the final gate.

No 15
Setup
Webflow advice for beginners

Three things I wish someone had told me when I started Webflow eight years ago: use Relume for layouts, follow Client-First for class naming, and switch to REM units.