Quick heads up, because there's a lot of panic flying about on this. Anthropic extended Fable 5 on paid plans through Sunday the 12th of July. After that it stays in the app, but it bills prepaid credits at $10 per million tokens in and $50 out — the most expensive model they sell. So everyone is racing to squeeze demos out of it before the deadline. And the smartest thing I've seen people do this week is the opposite: they're making Fable build the things that keep working after it's gone. The logic is simple. Claude reads your CLAUDE.md and your skill files before every session, no matter which model is running. So if the smartest model available writes those files, every cheaper model you run afterwards inherits that thinking. And those files don't expire on Sunday. Here's the prompt — run it in Claude Code before the weekend is out: "Rewrite my CLAUDE.md with everything you know about how I work. Go through this repo and our project history: how I like changes made, what I always ask for, the corrections I have made to your work, my naming conventions, and the tools I use. Write it as instructions to a future model that has never met me, so the next session starts already knowing all of it. Show me the draft before saving it." And if you want to go a level further, have it write skill files for the jobs you run most often — the steps, the quality bar, the checks it would run before calling something done, and the traps a smaller model would fall into on that job. Two more things worth doing before Monday: 1. Set a spending cap in your usage settings. The default on credits is unlimited, and that's how people end up billed silly money. 2. Save Fable-on-credits for the calls actually worth paying for — big specs, plans, and decisions you can't easily walk back. Your everyday models will be carrying the skills it wrote for them. This was actually build five of five in the video attached — five Fable builds ranked by how fast they pay you back, each with a receipt from someone already making money with it: a $400k website flow, a $7k/month content loop, and a Google Ads media buyer running over MCP. The other four are in there if you want them.

who can help me with this problem im haveing i dont understand i did what i could do and i can solve this problem i have already solved all my issues

Hello! New to n8n and glad I stumbled upon Nate's Ultimate Assistant video. Great stuff. I tried to follow along but have run into an issue in the 'Send Email' node. In the video I see 3 parameters passed - emailAddress, subject and emailBody. However in my local instance I only have emailBody. How do I get the AI Agent to pass 'emailAddress' and 'subject'? Thanks for any feedback. Have been pulling out my hair on this one.