The database I could see wasn't the database that mattered The AI tooling saw a database with the right name, right schema, plausible data — and production read a different one. Verify the target before you let AI write.
I set up a full subscription business without opening the Stripe dashboard Products, prices, webhooks, portal and a Swiss VAT registration — all over the Stripe API after the dashboard and connector both failed. And why billing is rented, never built.
The AI kept inventing football crests. The fix was a sentence. Eighteen AI-generated ad images, eleven redos: hallucinated trademarks, drifting product geometry, and the QA ruleset that made an image model production-ready.
Claude Cowork said nobody replied. The reply was message 28. An AI agent told me a thread had gone quiet. The reply was message 28 of 28, hidden by a truncated search. Why AI tools fail where they fetch reality, and the one-line rule that fixes it.
Why your affiliate applications keep getting rejected My affiliate applications kept bouncing with the same reason attached. The rejections were logical — I'd made one structural mistake. Here's the fix.
Your analytics dashboard is lying to you by leaving things out A brand I work with was sure it mostly sold bundles. The dashboard couldn't prove it either way — so I went and got the data it was quietly leaving out. The number ended the debate.
I pointed the Meta MCP at a real ad account. Here's what it found in ten minutes. A few weeks after wiring Claude into Meta Ads Manager, I ran a real ROAS audit on a live account. One zombie campaign, a hidden optimisation trap, and a lesson about where the human judgement still lives.
Meta just opened Ads Manager to Claude — here's the setup that actually works Meta just shipped an official MCP server that lets Claude, ChatGPT, and Gemini drive your ad accounts through natural language. Here's what it does, how to wire it up in five minutes, and what's quietly waiting to bite you if you do this without knowing the platform.