His Claude Code Workflow Is Insane

Summary

In this video, the creator of Cloud Code, Boris Chenry, shares his innovative workflow for coding, using multiple cloud instances and automating tasks through powerful integrations. By leveraging tools like Cloud Code's teleport, subagents, and MCPs, Boris demonstrates how he maximizes efficiency while maintaining accuracy and collaboration across teams. He emphasizes the importance of automation, AI validation, and constant context switching to streamline the development process. The video highlights how embracing AI tools can evolve your coding practices, save time, and enhance productivity for both individual developers and teams.

Takeaways

• 😀 Cloud Code is highly customizable, and each user may personalize it based on their workflow. Boris Chenry prefers a simple, 'vanilla' setup but emphasizes the flexibility of the tool.
• 😀 Running multiple cloud instances in parallel is an effective way to manage tasks efficiently. Boris runs five clouds at once and uses notifications for task management.
• 😀 Web cloud integrations with local setups can enhance productivity. Boris uses Cloud Code’s teleport feature to switch between local and web sessions seamlessly.
• 😀 The 'Opus' model (4.5) is Boris's go-to for coding tasks due to its superior tool-use abilities, despite being slower than smaller models like 'Snonit.'
• 😀 Cloud Code users can share a unified cloud.md for the team, allowing everyone to contribute to improving the system by adding findings and bugs.
• 😀 Integrating Claude into code reviews can help streamline the process, allowing automatic feedback and ensuring the team is aligned with best practices.
• 😀 Starting tasks in 'plan mode' helps ensure a solid foundation before coding. This step allows for deep analysis and thoughtful orchestration before execution.
• 😀 Automating repetitive tasks with slash commands (e.g., slash commit-push-PR) can save time and streamline everyday coding tasks.
• 😀 Using custom subagents, like Code Simplifier or Verify App, enables teams to automate common workflows and reduce manual effort in code reviews and testing.
• 😀 Cloud Code’s ability to integrate with external tools and systems (like Slack, BigQuery, or Sentry) enhances overall productivity, enabling automated reports and interactions.
• 😀 Long-running tasks can be managed through background agents or Ralph Wiggin plugins, avoiding unnecessary permission prompts while automating long-term processes.
• 😀 To get the best results, Cloud Code relies on self-verification. Running end-to-end tests and using agentic validation ensures code quality is consistently high.

Q & A

What is Boris Chenry's approach to using Cloud Code, and how does it differ from traditional workflows?

Boris Chenry emphasizes the flexibility of Cloud Code, showing that it can be customized to fit different workflows. He runs multiple cloud instances in parallel, utilizing a combination of local and web-based cloud sessions, which contrasts with traditional manual coding where developers typically work on one task at a time.

What is the benefit of using multiple Cloud Code instances in parallel?

Running multiple Cloud Code instances in parallel allows Boris to work on different tasks simultaneously without switching context too often. It helps him manage different aspects of a project efficiently, from coding to managing tasks like updating backlogs or fixing bugs.

How does Boris use the 'teleport' feature in Cloud Code?

The 'teleport' feature in Cloud Code allows Boris to switch between local and web cloud instances seamlessly. This feature helps him maintain continuity between devices, including checking progress on his phone throughout the day, enhancing the flexibility of his workflow.

What is the purpose of 'plan mode' in Boris's workflow, and how is it used?

Plan mode is used by Boris to carefully plan out changes to his codebase before making any modifications. By using plan mode, he ensures that he fully understands the context and scope of changes, which helps prevent errors and optimizes the final result.

What is the advantage of using Opus 4.5 in Boris’s workflow, despite it being slower than smaller models like Snonit?

Opus 4.5 is preferred by Boris because it offers better tool use and requires less steering, which, in the long run, can be more efficient than smaller models that might require more guidance. Despite being slower, its ability to handle complex tasks makes it a more productive choice overall.

What role do subagents play in Boris’s Cloud Code setup?

Subagents in Boris's setup automate common workflows, such as simplifying code or verifying app integrity. They help offload repetitive tasks and maintain context, making his workflow more efficient and allowing him to focus on more complex aspects of development.

How does Boris use 'slash commands' in his daily coding tasks?

Boris utilizes 'slash commands' for frequent tasks, such as committing code or pushing pull requests. These commands are stored in the .cloud/commands folder and allow him to execute commands more efficiently, saving time by reducing the need for repeated prompts.

Why does Boris recommend maintaining a shared 'cloud.md' file, and how is it used?

The shared 'cloud.md' file is used to document common issues, tips, and configuration details for Cloud Code within a team. It is checked into Git and updated regularly by team members, providing a reference for troubleshooting and ensuring everyone is aligned with the best practices for using Cloud Code.

How does Cloud Code's agentic validation improve the testing process?

Agentic validation allows Cloud Code to perform end-to-end tests on applications automatically. It simulates user interactions, takes screenshots, and reports back on the test results, providing confidence that the code works as expected without requiring manual testing.

What is the significance of using Cloud Code’s MCP (Managed Cloud Plugins) in Boris’s workflow?

MCPs in Boris's workflow integrate external tools like Slack or BigQuery into Cloud Code, enabling the automation of tasks such as posting updates, querying databases, or interacting with other systems. While MCPs offer powerful functionality, Boris advises using them cautiously due to potential risks like context window overload or prompt injection.