Tool-Specific Considerations

Tools used at Axelerant

Irrespective of the tool, we believe that using AI in right mindset is of paramount importance. Kindly go through this video on same.

We use ChatGPT for Enterprise/Codex and use Cursor for IDE based assistants for developers.

When to use what?

This guide helps engineering teams choose between ChatGPT, Codex, and IDE-integrated AI tools like Cursor, based on the task at hand. These tools serve different purposes across the development lifecycle—from ideation to deployment—and even during onboarding.

💬 ChatGPT (Conversational AI with Custom Context)

Best For:

• General-purpose exploration: architectural decisions, documentation, and brainstorming.

• Learning and onboarding: explaining code snippets, frameworks, or tools.

• Design support: outlining ADRs, evaluating trade-offs, and discussing patterns.

Usage Examples:

• “Explain how a decoupled Drupal setup works.”

• “What are the trade-offs between GraphQL and REST APIs?”

• “Draft documentation for our technical debt workflow.”

🤖 Codex (Code-as-a-Partner Automation Engine)

Best For:

• Natural language to code translation: creating usable code from English instructions.

• End-to-end automation: generating working prototypes, boilerplate, or PR-ready code.

• Cross-language support: converting logic across programming languages.

• Hands-on onboarding support for new engineers.

Onboarding Use Case:

• New developers can ask Codex questions about the codebase, such as:

  • “What does this function do?”

  • “Show me how to add a new endpoint.”

  • “Where is the authentication logic implemented?”

• Codex can generate example code that mirrors existing patterns in the project.

• It reduces onboarding time by bridging the gap between documentation and implementation.

Other Use Cases:

• Generate simple web apps or APIs

• Write SQL queries, data pipelines, or CLI scripts

• Automate routine code transformations

Advantages:

• Enables self-directed, code-driven learning

• Accelerates ramp-up with guided examples

• Scales across domains (front-end, back-end, DevOps)

🛠️ IDE-Based AI (Cursor)

Best For:

• Context-aware edits: makes precise changes in your local project based on full file visibility.

• Interactive coding: refactoring, code suggestions, real-time enhancements.

• Post-PR refinement: review and improve PRs generated by Codex or team members.

Usage Examples:

• Inline documentation

• Test generation

• Performance improvements and edge case handling

Short demo:

https://www.youtube.com/watch?v=oWz-0E5WzC4

Model/Tool Learnings:

Claude Sonnet: Often preferred for high quality, optimized outputs, and complex tasks, with fewer API calls. It's considered expensive but may justify the cost for quality and time efficiency.

Gemini: Provides a balance of cost and functionality. It may require more iterations or produce slower results compared to Claude, and can sometimes hallucinate.

Cursor AI: Strong for React/Node.js projects, understanding project context, and handling vague instructions. Offers strong design capabilities and agentic features

🧭 Decision Flow

Here’s a quick decision guide to choose the right tool:

References

• https://openai.com/index/introducing-codex/

• https://docs.cursor.com/context/rules