How Claude Code’s Creator Starts EVERY Project

Boris Churnney's Approach to Planning in Claude Code 00:24

"Probably 80% of my sessions I start in plan mode. Once the plan is good, it just stays on track and it'll do the thing exactly right almost every time."

• Boris Churnney emphasizes the importance of starting projects in "plan mode," stating that it accounts for 80% of his sessions. Planning effectively helps ensure that the subsequent building phase progresses smoothly and correctly.

• Many users tend to jump straight into coding without sufficient planning, which can lead to misunderstandings between what the AI thinks it is solving and what the user actually intends.

• Inadequate requests can cause AI to return incorrect solutions, resulting in time wasted debugging problems that could have been resolved early with a clear plan.

• Boris urges users to take time upfront for thorough planning, comparing this practice to the Navy Seals' philosophy of "move slow to move fast." He advocates for preparing the groundwork before diving into execution, leading to a more efficient process.

Utilizing the Claude.md File for Effective Maintenance 02:40

"If you hit this, my recommendation would be to delete your Claude.md and just start fresh."

• The Claude.md file serves as a personalized instruction manual for the AI and should be kept concise. Boris recommends that if this file becomes too lengthy and complex, it may be more beneficial to start over completely rather than trying to refine it gradually.

• He believes that as AI models improve, the requirements from six months ago may no longer apply and can be simplified in the current context.

• Users often make the mistake of over-adding to their Claude.md files, thinking that more instructions equal better performance; however, this can lead to confusion and decreased effectiveness.

• An alternative approach is to clean up the file periodically instead of deleting it entirely, ensuring that it only contains necessary rules that optimize the AI's performance.

Importance of Verification in Claude Code Projects 05:24

"Give Claude a way to verify its work. If Claude has that feedback loop, it will 2 to 3 times the quality of the final result."

• Verification is a crucial step in Boris's workflow where he encourages users to implement a feedback loop that allows Claude to check its own output for accuracy.

• To practically apply verification, users should give Claude a tool to assess the output and clearly communicate this to the AI.

• For coding tasks, this might involve testing built code in a real browser to verify its functionality, while content creation could focus on aligning outputs with established brand guidelines.

• Before commencing work, it's beneficial to hint at verifying the results in the Claude.md file, facilitating a systematic review of what the AI produces and ensuring overall project quality.

Leveraging Parallel Workflows for Increased Efficiency 07:00

"What you want is to do the minimal possible thing in order to get the model on track."

• Boris champions the idea of working in parallel by partitioning different workflows, which can dramatically enhance productivity when using Claude Code.

• By maintaining multiple Claude sessions running simultaneously, users can streamline their efforts and tackle various aspects of a project concurrently, maximizing the effectiveness of their time and resources.

• Understanding the balance between efficiency and over-complication is key; Boris advises to only expand on instructions when absolutely necessary, allowing the AI to adjust as required.

Partitioning Tasks for Efficiency 07:21

"Two context windows that don't know about each other tend to get better results."

• Boris emphasizes the importance of partitioning tasks to avoid overlapping efforts. Having multiple sessions working on the same task can lead to counterproductive outcomes, similar to two people working on the same problem at the same time.

• By creating separate sessions, each can approach a problem with a fresh perspective, untethered by the context of previous discussions. This can result in solutions that might be overlooked when focused on the same context too deeply.

Utilizing Inner Loops for Repeated Tasks 08:07

"I use slash commands for every inner loop workflow that I end up doing many times a day."

• Boris refers to repeated tasks as his "inner loop," and uses slash commands to streamline these workflows. This systematic approach allows him to execute frequent actions quickly without having to re-enter prompts each time.

• The concept of "Claude skills" can be utilized to document these repeatable processes, ensuring they can be executed consistently, similar to how players run specific plays in a game.

Building Skills for Consistency 08:40

"Think about the inner loops that you have throughout the day."

• Boris illustrates the application of Claude skills by using the example of generating a consistent report format. By developing these skills, tasks can be executed efficiently with only the input data changing from one instance to the next.

• Creating these skills leads to an infinitely repeatable process, enabling immense productivity gains in routine tasks.

Planning for the Future of AI Models 09:55

"The more general the model, the more specific the output will be."

• Boris points out the significance of building for the future, noting that AI models are continuously improving. He advises against spending excessive effort on optimizing prompts since the models will naturally enhance over time.

• Instead, focusing on how you provide information to the model and optimizing the system that surrounds it will yield better long-term results. This approach encourages users to adapt their workflows by anticipating the evolution of AI capabilities.

Key Takeaways from Boris's Workflow Strategy 11:30

"Document once, run forever."

• Boris presents several strategies that encapsulate his approach: use plan mode to think through discussions, create minimal documentation, ensure verification processes, utilize parallel sessions for problem-solving, and build for future improvements.

• This systematic and forward-thinking approach to workflow can lead to increased efficiency and productivity when working with AI tools like Claude.