Vous gaspillez 90% de vos tokens sans le savoir

Are you reaching your Claude limit too quickly? The problem isn't your subscription, it's how you're using your tokens. → 🔗 My step-by-step tutorial for setting up the Claude Code status line https://shubham-sharma.fr/ressources/... → ✉️ Join my newsletter https://shubham-sharma.fr/newsletter/ → 📅 Sign up for the waiting list for my private workshops https://sharma.fr/masterclass-privees What you'll learn in this video is why Claude sometimes blocks you after only a few hours of work, and most importantly, how to avoid burning through your tokens without even realizing it. We begin by understanding the real difference between message limits and token limits, then we see why overly long conversations make Claude more expensive, slower, and sometimes even less intelligent. Next, we move on to the 9 pitfalls that cause your consumption to skyrocket: conversations that drag on, overly verbose replies, poor model selection, unnecessarily loaded MCPs, an bloated Claude.md file, an invalidated cache, excessively large files, images, and misused sub-agents. We conclude with concrete best practices for Claude Code and Classic Claude: clear slashes, compact slashes, context slashes, status lines, choosing the right model, and a simple method for working all day without getting blocked at the worst possible moment. 00:00 Introduction 00:31 How Tokens Work 01:22 Conversation Hygiene 02:52 Poorly Managed Compact 03:32 Verbose Responses 05:03 Choosing Templates 06:21 Managing MCPs 07:36 Optimizing cloud.md 08:19 Caching 09:26 Files 10:34 Using Sub-Agents 11:31 Subscriptions and APIs 13:00 Conclusion If you want to learn more, feel free to follow me on my social media. I share daily tips on AI, Claude Code, and the workflows I use every day. MY OTHER ACCOUNTS: 🌍 My website - https://www.sharma.fr/ 👨‍💻 LinkedIn -   / sharmashubham1   📱 TikTok -   / shubhamnocode   🐦 Twitter -   / shubham_nocode   📸 Instagram -   / shubham_nocode