Practical AI - ARSTECH

Magnifying glass over a grid of blog post cards, with orange tabs flagging missing metadata and two duplicate cards.

Practical AI

Ghost MCP server: I let Claude audit my blog, and it found what I'd missed

I built an MCP server that gives Claude the whole Ghost Admin API. The first real job I gave it was an SEO audit, and it found missing meta descriptions, empty alt text and duplicate tags I'd never noticed.

Laptop with a code editor and AI chat, wired to two speed gauges reading 110 and 26 tokens per second.

Practical AI

Flint vs Qwen 3.8 27B on a MacBook Pro M5: my local AI setup in VS Code

I tried a pile of local models on my MacBook Pro M5 and kept two: Flint for fast fixes at 110 tokens/s, Qwen 3.8 27B for the harder work at 26. Here is the setup, the numbers, and how I split the work between them.

Developer at a laptop with a local AI chat beside a small home server and a cloud AI chat on one screen

Practical AI

Local AI can replace Claude for a lot of work, but not the work that scares me

A local AI model can handle more real work than most people expect. The problem is knowing which 80% is safe to keep local — and which 20% still needs a stronger model.

Developer studying a Supabase Row Level Security dashboard with table policies on a large monitor at night

Practical AI

Supabase RLS checklist for AI-built apps: stop leaking user data before launch

Supabase makes it easy to build fast, but a public anon key with weak RLS can expose real user data. Use this checklist before launching an AI-built app.

Practical AI

Stripe webhook security for AI-built apps: don't trust payment events without verification

Stripe Checkout can be added quickly to an AI-built app, but a weak webhook can mark users as paid without a real payment. Check these webhook security basics before launch.

Developer at a multi-monitor desk with a security checklist, vulnerability scan results and a launch readiness gauge

AI App Security

AI-built app security checklist: don't launch before checking these 15 things

AI-built apps can look finished while they are still leaking keys, exposing database rows, or trusting unsafe webhooks. Use this checklist before you put real users, client data, or payments into production.