The well-informed space

The well-informed space is the knowledge base that every cycle of the loops adds to. In the 2011 deck it was called the Knowledge Base/Portal. Its job was always the same: give a team a place to be curious, so the creative process can thrive on what is already known instead of guesswork.

What goes in it

From the 2011 and 2013 decks:

• Observations
• Problem articulations
• Narratives, scenarios and user journeys
• Multi-sensory artifacts (sketches, prototypes, recordings)
• Recommendations and proposals
• Reporting, including KPI reporting against ROI, quantitative and qualitative

What changed

For years the missing piece was retrieval. The space could hold everything, but nobody could find it again when it mattered. Large language models changed that. The February 2024 introduction put it this way: "With the advent of generative AI and large language models, the repository of potential has become an active, accessible wellspring of inspiration."

The next step is a space agents can browse and keep current, not just one people can search.

How to keep it from becoming a dump

The documented failure is dumping everything in raw and asking a model to find the themes afterwards. The shape that works:

• Synthesize per source first, then across sources.
• Add metadata when a record is written, not later.
• Give every record a date, a source, a confidence level and a review date.

This site works that way. Every page here is a Markdown file with the same front matter: type, loop stage, domains, status, version, sources and confidence. You can read the pages, or point an agent at llms.txt and the JSON feed and it reads the same records.

Open questions

• Is the well-informed space part of the DRF, or the substrate underneath it? The harness framing leans toward underneath.