You've been using AI Wrong

About this video

The creator profiles Fabric, an open-source CLI framework from creator Daniel, pitched as augmenting humans with AI by reducing friction rather than replacing chat interfaces. Fabric takes text from any source and sends it to a chosen model — OpenAI, Anthropic, or local/remote models via Ollama — using reusable, crowdsourced and inspectable prompts called patterns. He demonstrates piping a YouTube transcript obtained with a transcript tool built into Fabric into the extract_wisdom pattern to get summaries, ideas and quotes from a two-hour interview, and baking a custom workout_summary pattern into a Python script that processes Strava JSON. Setup shown is cloning from GitHub, installing with pipx, running fabric –setup to add OpenAI, Anthropic and YouTube API keys, listing models and patterns, changing the default model, and using -s to stream and -p to select a pattern. He also shows stitching patterns together (summarize then write_essay), creating a custom pattern as a system.md file under ~/.config/fabric in a separate custom folder so updates do not overwrite it, using a context file, and saving output directly to Obsidian.

Key takeaways

• Fabric is a CLI framework, not a model itself: it routes text to OpenAI, Anthropic or Ollama models using community-curated patterns.
• Patterns are open, visible system prompts for specific jobs such as extract_wisdom, summarize, analyze_claims, label_and_rate and improve_prompt.
• Piping is the core workflow: e.g. a YouTube transcript piped into fabric with -sp extract_wisdom, or one pattern’s output stitched into another.
• Install via GitHub clone and pipx, then run the Fabric setup command for API keys; Fabric also offers commands to list models, list patterns and change the default model.
• Custom patterns live as a named folder containing system.md under the Fabric config; keep them in a separate folder and copy them in so pattern updates do not delete them.

Episode details: published May 28, 2024 · 30:58 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are NetworkChuck’s own claims, shared for education — they are not typical or guaranteed.

← Previous: I use this everyday · Next: RetroPie: A Raspberry Pi Gaming Machine →

All NetworkChuck videos · NetworkChuck profile · NetworkChuck on YouTube