How I Run ZeroClaw on Modest Hardware
ZeroClaw makes a capable agent practical on modest hardware. My setup adds narrow tools, explicit confirmation for consequential actions, and local retrieval. The interesting …
An Agent Bus for My Coding Agents: MQTT, MCP and the Hard Part of Waking a Session
A small agent bus for coding agents spread across a home LAN: Mosquitto for presence, queues and routing, and one Rust binary that every agent starts as an MCP server. Updated after two days of real use: file attachments, short aliases, duplicate-id …
Archive
Measuring a Classifier at Scale: LLM Labels, Human Review
How to measure a production classifier when there is far too much traffic to label by hand: a blind LLM labeller does the volume, a verified replay attributes the errors, honest …
PostCraft: Turning an LLM Writing Pipeline Into Something I Could Actually Trust
How I used deterministic contracts and LLM-backed self-evals to make PostCraft generate grounded technical portfolio posts without inflating what the code actually supports.
From Embeddings to Ingress: The Complete Architecture Behind a Production RAG System
Arch Review is a personal engineering project that combines a RAG application with production-style platform concerns: ingestion, chunking, embeddings, retrieval, citations, a …
Designing a Multi-Environment Django Platform on Kind with Pulumi, Components, and Policy as Code
How I turned a simple local Django-on-Kubernetes deployment into a small but disciplined infrastructure codebase with Pulumi stacks, reusable components, validated settings, tests, …
Designing a Token-Efficient Personal News Podcast with MCP and LLM-Driven Enrichment
How I designed Better ArchNews as a layered Python system that enriches technical articles with LLM intelligence, exposes them through MCP for token-efficient retrieval, and turns …
Designing a Self-Hosted Backup Orchestrator with Go, Redis, and Rsync
JustBackup is a self-hosted backup orchestrator that keeps the data path boring (SSH + rsync) and focuses on operability; scheduling, async execution, restore workflows, hooks, …
How I Turned My Job Search Into an AI-Assisted Positioning System
I built a lightweight pipeline to collect job descriptions, normalize them into markdown, analyze them with LangChain and a local LLM, and score them against my CV. The goal is not …
How I Structure a Fraud Detection Side Project So Iteration Stays Cheap
A fraud detection side project became a useful engineering exercise once the codebase was structured around modular features, explicit pipelines, and repeatable evaluation.