Vibe Coding With Desktop Commander: Code Documentation + Obsidian + GitHub

Key Concepts

• Desktop Commander: A server providing tools to AI clients (LLMs) for interacting with a computer's file system, terminal, and processes.
• MCP (Model Context Protocol): A protocol that allows LLMs to access and utilize external tools and information.
• LLM (Large Language Model): An AI model trained on vast amounts of text data, capable of generating human-like text, code, and more. Examples include Claude, Gemini, and others.
• Obsidian: A note-taking and knowledge management application that uses markdown files and supports linking between notes.
• Context 7: An MCP that pulls up-to-date code documentation.
• GitHub: A web-based platform for version control and collaboration, commonly used for software development.
• Vault (Obsidian): A folder containing a collection of Obsidian notes and settings.
• Relative Links: Links in markdown files that specify the location of a file relative to the current file, allowing them to work correctly on GitHub.
• Double Bracket Links: A linking syntax specific to Obsidian, which does not work correctly on GitHub.

Desktop Commander Introduction and Features

• Desktop Commander is a product with a set of tools that gives LLMs access to your computer in an agentic mode.
• It allows LLMs to interact with the terminal, files, and processes, enabling tasks like code creation, landing page generation, and automation.
• Desktop Commander is free to use; users pay for the token usage of the LLM they connect to it (e.g., Claude).
• The presenters launched a new release with improved file search and onboarding.
• Example: Creating a shopping list app using Desktop Commander and Claude.
  • The LLM creates a folder and writes the necessary files for the application.
  • The application is then opened in a browser.
  • The presenters then publish the shopping list app to a new repository on GitHub.
• Desktop Commander can be used with various LLMs, and the presenters are working on their own client to support different models (including local models in the future).
• The new release includes improved search functionality, combining file name and text search into one tool.

Building and Deploying an Application

• The presenters demonstrate building and deploying a shopping list application in approximately five minutes using Desktop Commander.
• The application is a static application without a backend, but it can be extended with services like Firebase or Supabase.
• The presenters emphasize that building applications is becoming easier, and the focus is shifting towards finding clients.

Discussion on the Role of AI in Programming

• The presenters discuss the evolving role of AI in programming, comparing it to the printing press, which expanded the number of writers and readers.
• They believe AI can handle tasks like typo correction and documentation, allowing engineers to focus on higher-level engineering tasks.
• The presenters suggest that AI can act as a "junior engineer," handling routine tasks while senior engineers focus on architecture and design.
• The presenters are at the stage where Desktop Commander is building Desktop Commander.
• The presenters believe that no language is the programming language.

Integrating Desktop Commander with Obsidian for Codebase Documentation

• The presenters explore using Desktop Commander to manage knowledge and create documentation for codebases using Obsidian.
• They aim to create an Obsidian vault within a code repository and populate it with documentation for the codebase.
• They plan to use both Claude and Gemini (via their own client) to compare their performance in this task.
• The presenters choose the Context 7 repository as the target codebase.
• They use Desktop Commander to fork the Context 7 repository and create Obsidian vaults using both Claude and Gemini.
• Gemini completes the task faster, creating a basic documentation structure within the Obsidian vault.
• Claude takes longer but creates more comprehensive documentation, including architectural schemas.

Addressing Obsidian Link Compatibility with GitHub

• The presenters identify an issue with Obsidian's double bracket links, which do not work correctly on GitHub.
• They task Gemini with writing a script to convert these links to relative links that are compatible with both Obsidian and GitHub.
• Gemini successfully creates a Python script to perform the conversion.
• The presenters run the script and verify that the links are now compatible with both platforms.
• Claude also completes the task, but the presenters encounter conflicts when trying to push the changes.

Discussion on Code Comments and Documentation

• The presenters discuss the importance of code comments and documentation, particularly in the context of AI-driven development.
• They suggest that comments can provide context and explain the "why" behind the code, which may not be apparent from the code itself.
• They explore the idea of co-locating documentation with code to make it more accessible to AI models.
• The presenters discuss the layers of abstraction in software development, from low-level CPU instructions to high-level frameworks and documentation.
• They suggest that AI can enable a shift towards working at higher levels of abstraction, with engineers focusing on LLM documentation.

Addressing Audience Questions

• The presenters answer questions from the audience regarding the use of databases with Desktop Commander, the impact of AI on programmer jobs, and the choice of programming languages.
• They clarify that Desktop Commander uses the file system as its storage mechanism.
• They acknowledge that AI may reduce the need for large engineering teams but emphasize the potential for increased productivity and innovation.
• They discuss the reasons for using higher-level languages like JavaScript instead of assembly, highlighting the importance of abstraction and code reuse.

Conclusion

• The presenters summarize the demonstration, highlighting the ability to create and deploy applications and generate documentation using Desktop Commander and LLMs.
• They emphasize the potential of AI to assist with tasks like documentation and code generation, freeing up engineers to focus on higher-level tasks.
• They share links to the resources created during the stream, including the shopping app and the forked repositories with Obsidian documentation.
• They invite viewers to join their Discord server and sign up for early access to their client.
• The presenters thank the audience for their participation and announce plans for future live streams.