LLMs How to Use

[ ] I want to build my own screencast extension and linked to my GCS and the link can be rendered by markdown tag in my blog. [ ] 中外神话大战, global audience

you can watch the youtube video to learn how to use them:

• chatgpt: advanced data analysis

• claude artifacts: create web app etc and practice it and it can be published it to run on web for everyone access

  • produce flashcard on my english notes
  • artifacts base: https://claudeartifacts.com/?category=All
  • diagram generation: conceptual diagram
    • source code dependency diagram
    • book diagram, logic sequence

• cursor: composer, manage your local codebase, using claude remotely

  • pricing: 20$/month
  • can be used to wirte code in context
  • explain code for you, for example, from open source
  • compare with github copilot, price, feature, etc

• true voice mode in ChatGPT, not the vioce -> text tokens

• grok voice mode is interesting, many roles can play

• google notebooklm: read and extract

  • generate podcast

• image input

  • show in text to confirm the model see image correctly
  • In ChatGPT, you can paste the screenshot directly to it

• image output:

  • generate image about the big headlines today
  • provide the image from manhua first, and create your own story
    • see dall-e3: https://openai.com/index/dall-e-3/

• Video input:

  • maybe the model is still processing stream of images from the video

• Video output:

  • sora: https://openai.com/sora/
  • see X heather cooper comments comparison

• ChatGPT memory, you need to instruct it:

  • please remember it to our conversation context.
  • ChatGPT will store it in its memory bank about you, a separate DB.

• You can personalize the ChatGPT globally in settings.

• Create your own custom ChatGPT with instructions.

  • so LLM has the instructions as context to handle your questions
    • give LLM examples question/answer format, etc
    • for example: break korean into words, so easy to make flashcards later