Signals · in design with partners
How your organisation really works, with the working shown.
Every organisation has two versions of itself: the one on the org chart, and the one people actually live in. Signals listens to interviews and the systems your teams already use, and turns what it hears into conclusions you can check, question, and correct.
It keeps the world model honest. The design you set stays yours; Signals shows where reality agrees with it, where it doesn't, and how sure it is.
From a single answer to something you can rely on
Nothing jumps straight to a conclusion. Evidence earns its way up, one step at a time, and every step is recorded.
- 01
Signals
Raw evidence: an interview answer, a ticket, a roster change. Kept exactly as it arrived, with where and when it came from.
- 02
Facts
Plain claims a lens draws from a signal, like who owns a capability or which team another depends on.
- 03
Hypotheses
Statements about how the organisation works, gathering support or contradiction from facts over time.
- 04
Theorems
Hypotheses with enough independent evidence, over enough time, to be relied on.
Lenses: reasoning a person can read
A lens is a set of plain questions. Broad ones come first, and each answer decides which narrower questions get asked next.
A signal comes in
An interview answer, a ticket, a roster change, kept exactly as it arrived.
A broad question first
Which kinds of fact does it state? Every kind that applies opens its own branch, and the branches run side by side.
Then narrower ones
Each branch asks only what its fact needs: which team, relying on whom, for what. The options are your own teams, capabilities, and systems.
Every answer says how sure it is
Each question is answered by a small, fast model built for that one judgement. The answers it passed over stay on the record.
Sure enough, it becomes a fact
One signal can yield several, each with every step of its reasoning kept.
Unsure, a person decides
Never a guess. Questions that are often unsure are flagged for rewording.
Not in your map, it is flagged
A capability that already exists but isn't mapped yet is surfaced for you to add, not forced onto the nearest match.
Yours to change
Your people and practice leads can read a lens and change it without writing code. Every version is replayed against past evidence and your corrections before it goes live, so you see what it fixes and what it breaks.
Two kinds of finding, handled differently
"Platform owns deployment" and "Platform is under strain" are different kinds of statement, so Signals treats them differently.
Structure
Who owns what, and who depends on whom
Structural facts are checked against the design you hold in the world model, and shown as a traffic light that moves on its own as evidence arrives.
-
Green. Independent sources agree with the structure you have designed.
-
Amber. The evidence is thin or mixed. Worth another conversation.
-
Red. What people and systems report consistently differs from the design. That is drift, and you can see exactly why.
Signals never edits your design. When reality has moved, you decide whether to update the design or record the difference as intentional.
Conditions
Strain, friction, handoffs, and blockers
Conditions are a matter of degree, and they change. Their evidence fades with age, so a hypothesis reflects how things are now rather than how they were last year.
A condition never becomes a theorem automatically. It is put in front of people to discuss, support, or refute, and those conversations become part of the record.
Track a hypothesis over time, and reorganisations show up as clear turning points. You can see whether a restructure fixed a problem or just moved it to another team.
Every conclusion shows how it was reached
Click any theorem and walk back through it: the hypotheses that built it, the facts behind those, the lens that drew each fact, the questions it asked, the answer it chose, and the answers it nearly chose. Reasoning you can follow is reasoning you can challenge.
Correct a conclusion in place
If a lens got something wrong, anyone with access can fix that one conclusion on the spot. No one has to edit a lens to correct a fact.
Corrections teach the lens
Corrections pile up against the question that made the mistake. When a question keeps being corrected, it is flagged for tuning, and every past correction becomes a test the next version has to pass.
Evidence can raise new questions
Evidence that fits no existing hypothesis is kept. When enough of it, from enough independent sources, points the same way, Signals proposes a new hypothesis for people to discuss.
Where the evidence comes from
Interviews
Signals works alongside Metacosm's interviewer as the conversation happens. It tells the interviewer what it is still unsure of, so the next question is the useful one. It never tells anyone what someone else said.
Systems of work
Tickets, documents, and messages in the tools your teams already use show how work actually moves, rather than how anyone remembers it moving.
Your people systems
Rosters and reporting lines keep the structure current, so a reorganisation shows up as a reorganisation and not as a pile of contradictions.
Signals is built as its own service with open interfaces. It powers the Metacosm world model, and it is designed to run alongside your own data stack too.
Built inside the trust rules
Signals is the observation phase our trust rules were written for, and it is designed around them from the start. Findings about how the organisation works attach to teams, roles, and systems. Work itself is credited to the person who did it, so their skills have evidence behind them, and it is never used to rank people against each other. Nothing anyone says in an interview is shown as their words. No finding surfaces from a single signal. And raw evidence is held so it can be withdrawn cleanly, without breaking the record built on top of it.
Read the trust rules →Make the transition with us.
Metacosm is being built with a small number of Australian scale-ups making the move from AI-assisted to AI-native. Design partners shape the roadmap, get senior attention from both practitioners, and pay early-partner terms.