Knowledge that compounds.
Organise around the questions that matter, make your commitments explicit, and let what you learn change how you act.
For any team working towards goals whose pathways cannot be known in advance. It connects the questions you are holding, the assumptions guiding your work, the evidence you meet, and the commitments you act through.
Why it exists
What INO really means
Organisation shaped by inquiries.
Most organisations are explicit about what they will do. Plans, milestones and budgets get written down, and the learning is left to look after itself. Inquiry-native organising turns that around. What you make public are the questions you are holding, and the work you do is how you move them.
Organised around action
- What is explicit: promises, deliverables, milestones.
- What is implied: the learning, and the model of the problem underneath.
- Mobilises a coalition fast, because plans are familiar.
- Changing your mind costs you something, so the model is defended rather than revised.
Organised around inquiry
- What is explicit: the questions, the hypotheses, what would change your mind.
- What is implied: the action. While action happens, it's an instrument to move the question.
- Coherence holds while the problem keeps moving.
- Updating is the point, so criticism arrives as an input rather than a threat.
Both kinds of organisation act, and both learn. The difference is which one is written down where others can see it and argue with it. And a model made visible can compound.
Compared to a standard LLM chat, we have built a federated wiki with LLM-capabilities allowing knowledge across an organisation to compound.
What it lets a team do
More like a team member than a generic assistant
Everything you know, in one place that answers
Reports, notes, transcripts, calls, papers, the things colleagues sent you, and the research you've done. You can ask it to ingest any of this, and when it answers it will process the new information, compare it with knowledge already held, validate what needs validating, and compounds any new insights.
A shared picture that survives people being busy
Connecting the dots and cross-referencing across an entire team's work creates a shared picture, in ways that disconnected parts of the organisation or busy individuals could never do.
Proactive feedback, designed for your needs
The wiki can help you to assess new information against the information you already hold across the system, and inform you where there is discrepancy. It then helps a human to make sense of the discrepancy and reach a resolution.
It learns your work, in customisable ways
The routines it follows are written down and yours to change. For example, you can programme it to make a rule if the same correction happens three times, or build a skill for repetitive tasks. That is what makes it more like a team member rather than a generic assistant.
You talk to it. It does the filing.
No forms, no folders, no training course. You write a sentence or ask a question the way you normally would, and the work of connecting it to all the existing knowledge in the system happens in the background.
“Add this report.”
Filed. One summary, and updates to the six pages it touches: two organisations, three concepts, the overview. It contradicts what we held on outcome verification. I've recorded the contradiction and left it open for you. 7 pages · 1 contradiction declared · only you can see it, until you say otherwise
“What do we actually know about cooling in Madrid?”
Four sources, two syntheses, one open question. Here is the position, with each claim cited. Where the wiki is thin I've said so rather than filling the gap.
“Challenge that.”
Here is the strongest rival orthodoxy, argued as its own advocate would. If it lands, the position gets stronger. Filed as a challenge brief for you to decide on.
What comes out of it
Not a folder of documents. Things you can use.
A two-hour conversation nobody had time to write up becomes a position paper, with every claim citing the moment somebody said it. A tangle of risks becomes a map you can filter, grade by how well evidenced each link is, and walk through one step at a time. A set of goals becomes a view of which ones actually have somebody's time behind them.
These are not slides and not exports. They are live pages that rebuild themselves when the thinking underneath them changes, so a diagram you showed a funder in March still says what you believe in September, or tells you plainly that it no longer does.
And each one becomes material for the next. The position paper is what the next question is answered from, rather than starting again from the transcript. That is the whole of it: the work builds on the work.
How it's structured
Three interconnected layers.
Sources are loaded into individual wikis, which are private by default. At the discretion of the team member, key information can be pushed into the team wiki which is a maintained, interlinked set of pages. A third layer can be programmed with different rules and skills to pull structured information from the entire public repository, such as goals, commitments, and learning outcomes.
Sources
Reports, notes, papers, transcripts. The model reads them and can never change them.
The wiki
The interlinked knowledge repository layer.
Rules & skills
How the model behaves, and what it can do. Able to be programmed, and evolves over time.
What emerges
Structured information such as goals, commitments, asks, and learning outcomes – and the health of each. None of it is typed. Each is computed from the pages by a tool that only reads, then checked by a person.
One person, and everyone
Your own, and the team's.
Everyone has their own wiki. Teams have shared ones. Nothing moves between them until a person decides it does, and when it does it arrives saying where it came from.
That is what makes it safe to put real thinking in. Your half-formed work stays yours. What you send up becomes something the whole group can draw on, and what the group knows comes back down as a read-only copy your wiki can answer with but never quietly absorbs.
The shape grows with the organisation. Some links are deliberately one-way: a team holding sensitive material can read the centre and be unable to write to it, which is enforced by the tool rather than promised in a policy.
Designed to evolve
Interoperable and iterative at its heart.
Everything you collect stays yours, and the model is replaceable. Today it runs on Claude, but tomorrow if a better model existed or you wanted a change all of your knowledge remains because the wiki is an architecture of plain text files that can be read by any model.
The current system is currently being live prototyped by the Dark Matter Labs team. This means every day we are learning more, making changes, and evolving the system.
Now
A functional prototype
Adding sources, asking questions, health checks, a weekly catch-up, and ten wikis sharing with each other by consent. Working, in daily use.
Being built next
Something easier to use
An onboarding experience, improved user interface, enhanced workflows, and ability to draw from a wider variety of sources.
After that
Wikis that learn from each other
Improvements travel between organisations the way they now travel between colleagues: saying where they came from, by consent, and with nobody owning the centre.
In use at Dark Matter Labs
Counted on 21 September 2026, from the wikis themselves. How each figure is worked out, and how to reproduce it.
2 How knowledge flows
How we're working to build trust.
We know LLMs have limitations and we're working on building five layers to enable teams to use this technology without assuming everything it produces is correct nor exposing sensitive data. This is a work in progress, and here's an insight into where our thinking is currently at.
Read part two →Or start with us. This is not something handed over with a manual. We set your wiki up, sit down with you to work out what it is for and what it should push back on, and bring in your first sources together. One person, one folder, one week, and we are on the other end of it the whole time. Just email Gurden and we will find a time.