Working with AI. What actually helps.
Clean CLAUDE.md, well-tended memories, clear prompts — and the model still does whatever it wants. This collection isn't theoretical best practice. It's the tips that actually take friction out of our day — and the traps we sadly taught ourselves.
Inside the whitepaper
- Why CLAUDE.md, memories, and prompts are wishes — not rules
- The "1× hard + 1× easy" principle: working in parallel without breaking your focus
- Critical-System-Gate: the physical emergency brake for sensitive paths
- Tokenizer changes explained: same text, +35 % tokens — and what to do about it
- Prompt caching in practice — from 283 € to 47 € per month in a real-world setup
- Fast Mode = 6× standard price: when it pays off, when it burns money
- Spotting model personality shifts — and adjusting your own behavior accordingly
Who should read this
Request the whitepaper for free
We'll send the link by email.
Enter your name and email — you'll get the download link in a short welcome email. No newsletter, no spam, unsubscribe any time.
We can also review your AI workflow in person.
30 minutes, free of charge, no strings attached — we look over your team's shoulder and name the two or three spots with the biggest leverage.