Learn and Do with Peter Notes, 2024-04-26

Previous notes

Topics

  • 0h 30m, check-ins, questions, comments, idea
  • 1h 10m, Video Production and Post-Production
  • 1h 15m, Jankifiers
  • 1h 25m, Coding with LLMs
  • 1h 35m, Language Models and Language Model Managers

Questions, Comments, Ideas

  • What are some options for gathering a series of images I’ve created from Midjourney, and adding captions to each?

Goals

  • post to mighty networks?
  • kinda like a story

Next Steps

  • set up a shared Ghost or Substack or Gumroad shared art journal

Options

Questions / Comments

  • API for posting to MN?
  • file format conversion, GIF, MP4, WEBP, etc.
    • MN converts GIFs to .avif format
  • other platforms for longer form posts, stories, slideshows, etc.
    • video - tiktok, insta, reels
    • flickr
    • blog/email newsletters - ghost, substack, gumroad
    • smugmug
    • Apple Photos, Google Photos
    • Discourse forum
    • Discord?

Video Production and Post-Production

Coding with LLMs

  • RJ's Talking Slideshow and Sitting Timer
  • Web APIs
  • How big of a system can a non-developer create?
    • example: campground reservation system
    • modularization and hierarchy
    • frameworks
    • authentication and authorization
    • database
    • for web apps, "server"
    • Git and GitHub
  • What can be made easily with LLM code, and what can't?
  • What Web API's can supercharge a custom HTML app (like the Web Speech API)?
  • What ideas do we have for making apps with LLM code?
  • When does it make sense to use an LLM to create HTML apps vs. Python apps vs. Node apps, etc.?

Jankifiers

Lexiconstructor ideas

  • other languages (but have to worry about swears)
  • least commonly used words
  • https://en.wikipedia.org/wiki/Words_(Unix)

Next steps

  • Daniele and Pete make a Lexiconstructor with different word lists

Language Models and Language Model Managers

Pete's Goals for LLM Users

will need to be updated as the ecosystems evolve

  • powerful local LLM(s), easily managed
  • cheap access to GPT-4 or higher SOTA LLM
  • chat interface
  • easy export of conversations (ideally to Markdown)
  • ease of use
  • open source preferred

Other Benefits

  • no tracking or surveillance
  • conversations not used for training some company's AI