Ruben Hassid · 12 min read · Gen AI Tools
Prompting is the worst way to use Claude.
Read the original on ruben.substack.com →
The article argues that prompting Claude is the worst way to use the tool and introduces Claude Cowork, a folder-based workflow that treats your local files as a persistent 'Second Brain' Claude reads before every session. It claims Cowork can produce more tailored, higher-quality outputs and documents, but warns that setup and file-management can be a barrier for non-developers, and offers a free guide to implement the workflow.
Key Takeaways
- Prompt-based usage often yields generic results; providing context, tone, audience, and rules matters for quality.
- Claude Cowork and a local-file 'Second Brain' workflow can personalize AI behavior by using your own files as prompts that are read automatically.
- The approach can dramatically improve outputs (e.g., ready-to-use documents) but requires managing folders, MD files, and file formats.
- A well-structured project and file-organization strategy can scale AI usage beyond single prompts, leveraging reuse and consistency.
- There is a free guide to set up Claude Cowork and organize your Claude 'brain' for better results.
Excerpt
Stop prompting Claude. For example, this is the worst way to use Claude: If you don’t understand why this prompt is terrible, this free guide is for you. This will get you average results. And you will then blame Claude for it. But deep inside, you might know the real problem. You didn’t give Claude enough context. You didn’t explain your tone, your audience, your rules. Everything. But who wants to type 500 words of instructions just to get a first draft? Nobody does. So you don’t. And the output sounds like everyone else. Maybe some of you tried ‘Projects’…