Resource Gathering

General Refactoring Sub use cases:

When/Why we refactor:

Improve Code Readability
1. Makes the code easier to understand for current and future developers.
2. Uses clearer variable names, breaks up large functions, and simplifies logic.
Reduce Technical Debt
1. Cleans up "quick fixes" or poor design decisions made under time pressure.
2. Prevents future bugs and development slowdowns.
Make Code Easier to Test
1. Encourages separation of concerns and modularity.
2. Refactored code is often more unit-testable and reliable.
Enable Reuse and Scalability
1. Extracts duplicated logic into reusable components or utilities.
2. Makes it easier to extend features without breaking existing functionality.
Improve Performance
1. Identifies and removes inefficient logic, memory leaks, or unnecessary computations.
2. Not always the primary goal, but often a beneficial side effect.
Prepare for New Features
1. Refactoring before adding features ensures the architecture supports expansion.
2. Helps avoid “spaghetti code” when introducing complexity.
Fix Hidden Bugs or Prevent Future Ones
1. Cleaning up messy or unclear code can reveal logic errors.
2. Refactoring can proactively prevent bugs caused by unclear logic paths.

Sub-use cases of Refactoring:

Partial refactor and Recommendations
1. General refactor of parts of the application (for the reasons listed above)
Complete rewrite
1. to a new language
Update framework/dependency versions

Scale measures

Scale	Description

Scale	Description
0	Could not perform task at all
1	Could perform task with a lot of guiding, context, and errors. Does not follow best practices, has multiple errors that must be resolved. Has all of the following: errors, organization issues, lack cleanliness, etc.
2	Could perform task with a large amount assistance and guiding. Has errors and bugs. Has most of the following: errors, organization issues, lack cleanliness, etc.
3	Could perform task well with a fair amount of guiding. Has some errors/bugs, but they can be resolved somewhat easily. Has several of the following: errors, organization issues, lack cleanliness, etc.
4	Could perform task easily with some guiding but still has some outstanding issues. May have some of the following: errors, organization issues, lack cleanliness, etc, but not all of those issues exist.
5	Could preform task with minimal difficulty/guiding, follows best practices, organized code, and correct typing. May have one of the following: errors, organization issues, lack cleanliness, etc.

Tools

Tool	Description	Links	Rating

Tool	Description	Links	Rating
Copilot with Claude and GPT, has free tier	The VSCode addon we know and love. This is a great tool that has the ability to get the full context of the codebase since it is in VSCode	https://github.com/features/copilot	Partial Refactor/Recs: 5 Complete Rewrite: Version Updates: 5
Q Developer (CodeWhisperer) VSCode extension has free tier	Can be used in-line and has a right-click menu, similar to Copilot. Uses Claude. Can have rules added directly from the sidebar, and context adding is simple.	https://aws.amazon.com/q/developer/	Partial Refactor/Recs: 5 Complete Rewrite: 3 Version Updates: 4
SonarQube 14 day trial		https://www.sonarsource.com/products/sonarqube/	Partial Refactor/Recs: Complete Rewrite: Version Updates:
CodeBuddy VSCode extension Has voice response option		https://codebuddy.ca/	Partial Refactor/Recs: Complete Rewrite: Version Updates:
Cursor Trial period		https://cursor.com/en	Partial Refactor/Recs: Complete Rewrite: Version Updates:
Aider	Terminal based pair programming	https://aider.chat/	Partial Refactor/Recs: Complete Rewrite: Version Updates:
GPT-Migrate	Python app that runs in Docker.	https://medium.com/@fareedkhandev/gpt-migrate-convert-your-codebase-to-any-language-framework-cc1f846e4630	Complete Rewrite:
CLine		https://cline.bot/	Partial Refactor/Recs: Complete Rewrite: Version Updates:

Codebases for POC:

Staffing App
Status Reports
Angular Place My Order app?
React Native PMO?

Paths to consider

Measuring code quality with Code analysis tools
Identifying technical debt, and recommending changes
Custom GPT (There is already a devops one)
Complete rewrites: How to take a codebase in React and rewrite it in React Native (or insert two other frameworks here)
- For this, I think it’s important first to understand the codebase.
- 🧠✍️ Content idea: Understand a codebase with AI and then rewrite it with AI

Content for ideas:

https://www.youtube.com/watch?v=LsQGilvXAfE
Good for giving AI context before having it add new features https://www.youtube.com/watch?v=weaal3xZomc
- “Explore the entire repository to understand the codebase from multiple angles: as a software architect, software developer, and product manager. Compile your findings into a very extensive Markdown document in the root of the repository. For describing technical concepts, use Mermaid diagrams in the Markdown file.”
- Add the markdown file in the user > settings.json file.
- `
  "github.copilot.chat.codeGeneration.instructions": [ { "file": "CODEBASE_ANALYSIS.md" }, { "text": "Always add a comment: 'Generated by Copilot'." } ]
- “Act as a Product Manager for theStaffing App. We need an Advanced Analytics Dashboard page with visulizations (heat maps, skill demand trends, capacity forecasting), drill down capabilities (high level metrics to individual employee details), and export functionality (pdf reports, csv data exports). Do not code anything yet, instead, deeply think about which requirements we need and how the page should operate and describe the user stories in detail in a Product Requirements Document”
- fully implement the PRD as an MVP that, for now, just uses static frontend data.
Prompt formatting ideas, and parellel agents with Claude Code: https://github.com/disler/benchy/blob/main/specs/ui-revamp.md , https://www.youtube.com/watch?v=f8RnRuaxee8
Workflow sample: https://www.reddit.com/r/ChatGPTCoding/comments/1kxjtjm/my_ai_coding_workflow_thats_actually_working_not/
Workflow sample (CLine): https://www.reddit.com/r/CLine/comments/1kgikrs/help_with_best_practices_for_refactoring_code/
AI build sample (Potpie): https://www.reddit.com/r/AI_Agents/comments/1isimqr/i_built_an_ai_agent_that_makes_your_project/
Example of something WE DON’T WANT: https://www.youtube.com/watch?v=JWgyoX8ZCi8

General Refactoring

Approaches
- Basic approach without giving full context: “Refactor this component to improve readability” in Copilot
  - GPT4
    - Removes eslint comments, but does not fix missing dependency.
  - Claude Sonnet 4
    - ignores existing eslint comments to keep a dependency out of a useEffect unless explicitly asked to keep.
- The context builder:
  - Ask LLM to scan entire codebase, creating a Markdown document that explains the codebase in enough detail for a developer to understand.
  - Use that document for the LLM to reference as you begin prompting. The idea is that the LLM can reference the doc instead of the entire codebase each time.
Insights
- Copilot has greater context for the code as it’s an extension of VS Code.
- How to get LLMs appropriate context.
- Add rules for LLMs to follow while refactoring
Takeaways
- Copilot does a great job of taking reusability, separation of concerns, scalability, and readability into consideration.
- Web-based Claude creates separate files, which is a nice feature.
  - Code may be over-engineered.
  - ignores existing eslint comments to keep a dependency out of a useEffect unless explicitly asked to keep.
- Copilot can be good for adding features, and writing up user stories/PRD when given context.
- Need to make sure that with any refactor work, the LLM has sufficient context to complete the task. Workflows should include initial context set up, rules configuration, and THEN attempts at problem solving, followed by tuning/refining the results.