The Knowledge Brain: How to Turn Your Team's Dispersed Information into a Living Asset · LAB

Every day, a team generates meeting transcripts, interviews, internal notes, and documents that end up scattered across folders, chats, and different tools. That information exists, but it doesn't work for anyone.

On October 1st, at the Runroom LAB, Laura Polls (Head of Experience Research & Director of Runroom Academy) presented how we build and manage Knowledge Brains: shared knowledge bases, or organizational wikis, supported by Obsidian and Claude. The goal is to transform a team's or company's dispersed information from a passive asset into a living asset, structured and actionable for decision-making.

The Brain is a digital product

The first conclusion of the LAB: the most common mistake when starting a project of this type is to begin with the tools (Obsidian, Claude, NotebookLM).
The thesis is to start with the people and treat the Brain like any other digital product, with a Lean methodology (Lean Product Discovery and Management).

This implies defining three things before opening any software:

• Who it is for. A Brain can be individual, project-based, client-based, or organizational.
• What specific problem it solves.
• In which moments or work rituals of the team it will be consulted.

A Brain is also not designed perfectly from the start. It is built through hypotheses, starting with small cases and with a willingness to simplify, pivot, or discard structures when they do not add real value.

The ecosystem: Obsidian, GitHub, and Claude

Our way of structuring the Brain started from Andrej Karpathy's LLM wiki. With use, we moved away from that model and made our own decisions.

The architecture presented at the LAB combines three pieces: The architecture presented by Laura at the Runroom LAB combines three pieces:

• Obsidian as a Markdown viewer. It works with plain text files (.md), connects "knowledge atoms" through hyperlinks, and generates an interactive and visual network map.
• GitHub as a central repository. It synchronizes files among the team and securely maintains the change history.
• Claude as an AI assistant. It allows interaction with the Brain in natural language, building custom dashboards, executing advanced analyses (such as methodological maturity assessments), and resolving code or synchronization conflicts.

Four functions to make a Brain sustainable

For a knowledge base to function over time, it must be structured around four functions:

1. Nurture. Raw information input: meeting transcripts, interviews, internal notes, PDFs.
2. Organize (ingestion). Process and clean the data by extracting structured properties such as roles, categories, and contexts. This saves tokens in searches and effectively connects concepts.
3. Consult. Adapt access paths to each need: from direct readings to briefings, interactive visual dashboards, or automatic alerts in Slack.
4. Care. Manage permissions, entry criteria, review frequencies, and quality control to prevent the accumulation of obsolete information.

Learnings and conclusions

Standardize only if it adds real value. The team first tried to standardize meeting ingestion with a rigid folder structure (inbox and ingest). Upon detecting it generated friction, they pivoted to a model where each member does the ingestion individually, according to their need, before uploading the processed material to the shared Brain.

Strict traceability of information. The system must distinguish the raw source (the transcripts) from interpretations, solution hypotheses, and final decisions. AI should not present a hypothesis as a verified fact.

The human in the loop. Technology speeds up synthesis and structuring. Human judgment remains essential to validate the relevance, currency, and truthfulness of the knowledge.

Design through conversation. Creating a Brain does not start by opening software. It begins by debating the team's dynamics and real information consumption needs.

A Brain starts with a good question

The idea that sums up the LAB: before choosing tools, it's worth asking what decisions we want to make and what knowledge we need at hand to make them. Technology comes later and is adjusted with use.

Runroom LAB is a monthly in-person and free event. You need to register to reserve your spot. Now also available in streaming!