Dan Farrelly open-sourced Utah and it maps 1:1 to joelclaw. When two systems built independently arrive at the same design, pay attention.

The distinction isn't harness vs framework. It's between frameworks that own your execution model and frameworks that provide scaffolding for yours.

A spec for building durable agent memory. Give this page to a coding agent and ask it to build one.

A specimen report on personal AI infrastructure as biological organism. Permission scope as evolutionary axis. 187 architectural decisions deep.

A spec for building a feedback-driven AI writing system. Give this page to a coding agent and ask it to build one.

I replaced brittle bash watchdogs with a single Rust daemon that supervises the worker, probes the cluster, escalates failures, and actually tells me when things are on fire.

The people, books, and ideas behind the Knowledge Adventure Club - value paths, portfolio clubs, Understanding by Design, and the D&D metaphor.

Building system agents is like Minecraft. Blocks, logic, and an open world. The substrate is clay.

A pattern guide for building a self-hosted voice agent with LiveKit, a SIP trunk, and an LLM. Docker Compose or Kubernetes — the shape is the same. Written for agents building on behalf of humans.

Rob Pike and Ken Thompson spent the late 1980s asking what Unix would look like if you actually followed through on 'everything is a file.' Plan 9 was the answer. It failed commercially and succeeded intellectually — and its ideas now run the infrastructure we all depend on.

Joe Armstrong built fault-tolerant systems from intuition about processes and links. Phil Wadler proved that linear logic makes those same communication patterns deadlock-free by construction. I'm building an Elixir agent to test whether any of this matters in practice.

The slots pattern for Next.js cache components — pass ReactNode through cached boundaries to keep static shells fast while streaming personalized UI through Suspense holes.

Andrej Karpathy bought a Mac Mini to tinker with Claws. Simon Willison thinks 'Claw' is becoming a term of art. I've been building one for a week and today I coded while driving.

Building a janky voice interface for my AI assistant. It's not natural, the lag is real, and it needs tons of work — but it beats typing while walking.

Build a pi extension that sends Inngest events and shows live run progress in a persistent widget. The full walkthrough — from empty file to working tool.

Saša Jurić's 2019 GOTO talk is a masterclass in fault tolerance. It made me look hard at my current TypeScript + Inngest + Redis + K8s stack and ask: what if BEAM just... does all of this natively?

Your AI agent's primary interface isn't a chat window. It's stdout. Here's how to design CLIs that agents can actually use — JSON envelopes, HATEOAS navigation, NDJSON streaming, and zero guesswork.

From ephemeral terminal sessions to an always-on gateway. Event routing, heartbeat monitoring, failure detection, and Telegram access — for an AI agent running on your Mac.

Durable workflows on your Mac. Each step retries independently. Self-hosted, zero cloud dependencies. Your agent's nervous system.

Three containers and a spike that turned into a production migration. Why I moved my personal AI infrastructure from Docker Compose to k3d — and the gotchas nobody warns you about.

Every AI session on my machine starts from zero. The system knows who I am but not what happened yesterday. Here's how I'm building an observation pipeline to fix that — stolen shamelessly from four open-source projects and one very clever pattern from Mastra.

Notes on Sean Grove's talk at Sanity's Everything NYC 2025 — spec-driven development, the steam engine metaphor for token power, and why everyone becomes a leader when you have a million agents.

What happens when you use a social networking protocol as the foundation for a personal AI system? I'm not sure yet, but the identity and trust primitives are fascinating.

OpenClaw showed me what a personal AI system could look like. Now I'm building one from scratch to understand how it all works — not to ship a product, but because learning by doing is the only way I know.

Every pipeline in my personal AI system flows through Inngest — video ingest, transcript processing, content enrichment, autonomous coding loops. Here's the architecture and why durable workflows change everything.

Notes on the Lex Fridman interview with Peter Steinberger — the guy who built OpenClaw in three months, hit 180k GitHub stars, and accidentally created the most chaotic open source saga of 2026.