Stop copying our AI setup. Copy how we think.

Tom and Paco present a discussion about AI productivity systems, centered on the principle that while a solid foundational methodology is valuable, implementation should be customized to how each person thinks and works. They contrast their different approaches: Paco uses a terminal-based interface with dynamic HTML artifacts and relies heavily on outline-based thinking for writing and complex projects, while Tom has built an interactive visual system with deliverables folders, session canvases, and multi-view knowledge representations.

The conversation covers several key systems they've developed:

1. Deliverables Management: Tom demonstrates an interactive HTML-based deliverable system that tracks decisions, audit trails, media files, and agent interactions. This addresses the problem of information being scattered in chat histories by creating structured views of work progress.

2. Team Knowledge Visualization: Tom showcases multiple visualization approaches (index view, graph view, and connection view) for displaying guidelines, SOPs, and workstreams. This system automatically interconnects documents based on file structure without requiring additional metadata, revealing the complexity of their documentation (111 SOPs, 72 guidelines, 43 workstreams).

3. Conversation-Centered Workflows: Both speakers emphasize keeping conversations with AI central to their workflows. Paco highlights how he maintains focus by using terminal input with visual output on separate screens, while Tom demonstrates embedding chat interfaces directly within deliverable views.

4. Customization vs. Standard Implementation: The core message is that their IICOR methodology provides a framework, but users should adapt it to their thinking patterns. Paco's text-oriented, outline-based system differs significantly from Tom's visual, multi-view approach, yet both work effectively because they match individual workflows.

5. Local-First Architecture: Both speakers emphasize that their systems work entirely with local folder structures rather than databases, demonstrating that proper documentation and file organization can replace traditional database dependencies while remaining LLM-agnostic.

They conclude that understanding the philosophical approach behind systems is more valuable than direct copying, and that as AI capabilities improve, well-documented systems will naturally improve alongside them.