Ruben Hassid Advocates for a ‘Minimalist’ Claude Setup for Enhanced Productivity

In a recent LinkedIn post, Ruben Hassid shares his unconventional approach to configuring Claude, advocating for a stripped-down setup that he claims leads to significantly higher productivity. Hassid, who has spent 194 days refining his method, argues that many users overcomplicate their AI interactions by providing too much persistent context.

Hassid outlines a strategy focused on removing elements that he believes hinder creative thinking and lead to unintended context leakage. Central to his philosophy is the idea of invoking information only when necessary, rather than having it constantly accessible.

“My entire Claude settings are (completely) empty. It’s the most productive Claude I’ve ever used.”

Deconstructing the ‘Empty’ Claude Setup

Ruben Hassid details several key components he has removed from his Claude environment. These include the traditional folder system, an ‘about-me’ file that is instead converted into an invoked skill, and global instructions that provide excessive context. He also advocates for disabling most settings and prompt templates.

Eliminating Hindrances to Creativity

Hassid points out that persistent context, even with instructions to ignore it, can lead Claude to access information the user does not intend. He states:

“Even when I prompt ‘don’t check my folders,’ Claude still checked & got the context that I did not want it to get.”

Furthermore, Hassid argues that providing too much global context can stifle the AI’s ability to think creatively. Instead of a comprehensive ‘about-me’ file or broad instructions, he suggests turning these into skills that can be manually invoked. This selective approach, according to Hassid, ensures that the AI only accesses relevant information when prompted, preventing it from becoming rigid or biased by extraneous data.

The Power of Skills and Projects

In contrast to the deleted elements, Hassid highlights two core components that form the backbone of his productive Claude setup: Claude Skills and Projects.

Claude Skills: On-Demand Knowledge

Hassid emphasizes the utility of Claude Skills, which can be triggered by a simple “/” command. He explains that his ‘about-me’ file and other essential documents are now saved as skills. This means they are only brought into the conversation when explicitly needed, aligning with his rule: “write the same prompt twice? Make it a skill.” This approach ensures that the AI’s focus remains sharp and uncluttered by information that isn’t immediately relevant to the current task.

Projects: Isolated Contextual Memory

The second key element is the use of ‘Projects.’ Hassid describes this as creating a dedicated space for each client, repeated task, or campaign. Within a project, all interactions, emails, and past campaign data are remembered, but critically, this context does not leak into other unrelated chats. This isolation is crucial for maintaining accuracy and preventing cross-contamination of information between different clients or tasks.

“Every chat inside remembers everything – the emails, past campaigns. Nothing leaks into the rest of your future chats.”

Hassid concludes his post by providing a clear, step-by-step guide for users to download and implement his ready-made Claude Skills library, encouraging a simpler, more effective way to interact with AI tools.