General Tips
Provide clear objectives and direction
General
Project-scoped
Milestone-based
Instruction-based
Specialized Tasks
Building New Optimized Frameworks or Algorithms
Use multiple high-tier LLMs in common mainstream chat assistants such as Claude, ChatGPT, and Gemini for methodological research, conceptual brainstorming, design, and implementation planning.
Use mid-tier LLMs to generate coding plans as Product Manager or Quality Assurance.
Use another mid-tier LLM to generate the coding plan as the main Implementation Engineer, or
Use "unlimited", cheaper LLMs to generate the coding plan as the main Implementation Engineer.
Use evolutionary component research and development based on first principle thinking
Use high-tier LLMs to describe a list of foundational components and make an evolutionary plan
Begin by creating one folder for the most challenging or fundamental components, use high/mid/unlimited-tier LLMs to implement the plan, test and evaluate it until you achieve the needs
Do quality assurance for each evolution
If satisfied, copy the folder into the next evolution, do the next plan by integrating the secondary foundational or most challenging components, repeat the same process, and benchmark the current work against previous works
Use high-tier LLMs to design the coding interface for these frameworks or algorithms that balances user ergonomics and performance, and integrate it with the improved state
Keep repeating the steps until it covers the core part of the optimized frameworks or algorithms
Engaging UX/UI Design, Strategy, and Implementation
Use multiple high-tier LLMs in common mainstream chat assistants such as Claude, ChatGPT, and Gemini for methodological research, branding, conceptual brainstorming, marketing strategy, design systems, quick mockups, complex component generation, and coding guides.
Use mid-tier or cheaper LLMs to generate code as Implementation Engineers.