In a recent LinkedIn post, Ruben Hassid discusses how users can simplify their Claude setup and focus on the outcomes they want from the AI assistant. This article reports on Hassid’s recommendations and the reasoning he gives; it is coverage of his post, not an article written by him.
“the best Claude setup is no setup”
Hassid’s central argument is that elaborate configurations can get in the way of discovering what a model can do. He advises readers to clear out accumulated instructions, skills, memory settings and folders, then rebuild around a specific task. He also cautions that any configuration can become outdated, suggesting users revisit it after several months.
A leaner approach to configuration
According to Hassid, users should first review their account and organization settings, then choose Auto mode in Claude Chat. He presents this as a way to let Claude proceed without pausing for approval at each step. His post also recommends trying Claude’s desktop app and Claude Code, while acknowledging that these choices are part of his own suggested workflow.
Hassid argues that clearing existing customizations can make room for experimentation. His checklist includes removing old instructions and skills, disabling memory, and setting aside “any over-organised folder” a user may maintain. Rather than treating a detailed setup as a permanent advantage, he frames periodic resets as a chance to see how the model behaves with fewer constraints.
“Everything is code. A slide. An image. A video.”
Describe the goal, then choose the format
That claim underpins Hassid’s advice to use Claude for creating things, not only for describing how to make them. He points readers toward the Design and Slides features and suggests prompting iteratively, including by uploading visual references. The post’s broader message is that users should state what they want and what it is for, while allowing the assistant to work out an approach.
A prompt built around context
To help clarify a request, Hassid offers a template asking users to explain their goal, audience, existing materials and criteria for success. It also asks Claude to restate its understanding and request more context when needed. As Hassid notes, “It can’t guess what you want (and why).” His emphasis is on making the intended outcome legible before asking for a deliverable.
He further recommends requesting an HTML page when a web-based artifact would suit the task, rather than defaulting to a Word document, presentation or PDF. The recommendation reflects his preference for interactive outputs, though the post does not argue that HTML is suitable for every use.
Resetting as a recurring practice
Hassid says he plans to revisit the advice in five months and cites a separate remark attributed in his post to Boris Cherny, whom he identifies as a Claude Code builder: “Every 6 months, delete your Claude MD. Delete your skills. Delete your hooks.” Together, the suggestions present configuration as something to reassess over time. Hassid’s post ultimately encourages users to begin with a clear goal, test the available tools and remove setup that no longer serves their work.
📝 About This Content
This article is based on insights shared by Ruben Hassid on LinkedIn.
📅 Originally posted on November 4, 2026 | View original post on LinkedIn →