Writing 40 articles · 2023 → 2026

What I learn while building.

FREN
App Store Connect: submitting an iOS app from the terminalEvery submission, the same time lost clicking and uploading screenshots. This time: a bash script, the App Store Connect API and Chrome driven by Claude Code. One human click.Strix: an autonomous AI hacker to audit a vibe-coded appAn open-source AI pentest agent that attacks your code like a real hacker and proves each flaw with an exploit. Docker, LLM, Apache 2.0 - the tool I plug into my security audit.3D landing pages with Claude Code and Opus 5: the anti-slop workflowBuilding a Three.js landing page that stands out: the right reference, Opus 5 pushed to the max, a harness that drives the browser, then molding. A video summary.Tutorial: building a 3D landing page with Claude Code and Opus 5The full walkthrough of the DesignCode video, section by section with timestamps: desktop, references, Mobin MCP, minimal prompt, maximum effort, molding, textures, shaders, skills.A real LAMP stack on your Mac: leaving MAMP Pro for HomebrewApache, PHP-FPM, MySQL, .test vhosts, local HTTPS and MailHog: everything installed with Homebrew and driven by a single command. The complete tutorial for leaving MAMP Pro, written for my students.Fifteen student accounts in a single commandEvery September, two hours of clicking through cPanel to set up my students' hosting. SSH access, a command-line API, a skill tested and fixed - it now fits in one command.Taking over a WordPress site: from live to local over SSHA therapist's site on Combell, zero documentation. One SSH key, WP-CLI, rsync, two MCP plugins - an hour later the clone runs locally. The full workflow, pitfalls included.Loop engineering: stop prompting AI, design the loopPrompt engineering is about asking for a task well. Loop engineering is about building the system that pursues, verifies, and corrects that task.Putting Claude Code in a loop: /goal, /loop, and hooksClaude Code offers several ways to keep working without waiting for your next prompt. Here is what each mechanism actually does.A good loop is not AI left on its ownAn agent’s autonomy needs measurable criteria, persistent memory, limits, and human stop points.I built a graphical memory for all my projectsHow I installed Graphify, assembled a global graph of 12 projects and used Ollama to give images a place in my development context.Deploying Your App to the App Store: EAS Build vs FastlaneVibe coding has democratized app creation. But once the app is built, how do you get it onto the App Store? Two approaches depending on whether your app is React Native or native Swift.