A 35B local model, briefed per task, calling out what Claude misses. The cheapest correction I have found — and it runs on the laptop I already had.
I built a five-column kanban board so the agents and I could see the same picture of what is in flight. Smallest infrastructure I shipped this year.
Real numbers from a working AI newsroom — running costs, what failed, and the lessons that only show up after you ship.
Kinetic typography, code rain, and AI-generated illustration. How category-driven design keeps an automated blog from looking automated.
Three AI personas argue strategy at 2 AM while I sleep. How structured multi-perspective debate replaced the blank-slate problem with compounding insight.
How I built an automated AI content pipeline with Claude, ComfyUI, and Telegram. From topic research to one-tap publish — where human review is the design, not the bottleneck.
How I built a fully automated AI blog with Claude, ComfyUI, and a Telegram approval workflow — and why human oversight is the most important feature.
How I built a three-layer AI security system for a solo-operated VPS — using shell scripts, two Claude instances, and automated threat intelligence.
How a weekend of vibe coding with Claude turned a classroom frustration into a production Arabic learning app — with audio generation, flashcards, and A-grade security.
Practical lessons from building real projects with AI code generation — from environment hygiene and prompt design to security, testing, and knowing when to start over.
How the Doherty Threshold shapes user experience design — why systems that respond within 400 milliseconds keep users engaged, productive, and in flow.
How Postel's Law applies to user experience design — building interfaces that deliver consistent outputs while gracefully accepting the full range of human input.