Your Design System Can’t Fix This
Parallel Truth And The Illusion Of Consistency
Key observations
- Design systems achieve visual consistency but cannot reconcile divergent meanings of underlying concepts.
- "Parallel Truth" occurs when a single concept has multiple, unacknowledged authorities within a system, leading to an illusion of agreement.
- This semantic instability causes user hesitation, double-checking, and systemic drift, signalling a failure in governance over meaning.
- Atomic architectures and AI-generated surfaces amplify the problems of Parallel Truth by propagating ambiguity at scale.
- True consistency is the absence of Parallel Truth, demanding a single, central authority for each concept within a system.
Most organisations think they have consistency.
They have consistent buttons.
The badges align. The spacing is immaculate. The components behave impeccably. Somewhere, a design system is quietly congratulating itself.
Somewhere else a group of designers are doing just the same thing, but more stylishly and dressed in black.
And yet…
“Owner” means creator in one module and current responsibility in another.
“Status” is a state here and a stage there.
“Priority” shifts depending on which screen you are looking at.
The interface looks stable. The system underneath is negotiating reality. This is not sloppiness, it is something subtler.
It is what happens when the same concept has more than one authority inside a system.
I call it Parallel Truth.
Parallel Truth is not duplicated data. It is duplicated authority. It is when a concept - status, owner, approval, risk - is defined, enforced, or interpreted in multiple places, without a single, acknowledged home.
A coherent system has one authority per concept.
An incoherent system contains parallel truths.
And parallel truths create the most dangerous kind of instability:
The illusion that everyone agrees.
The Layer We Mistake For Agreement
Design systems govern surfaces.
They standardise components, tokens, interaction patterns.
They answer practical questions:
- How should this look?
- How should this behave?
They are excellent at that.
They do not answer:
- What is this, exactly?
That question lives beneath the interface, in the semantic foundation.
A semantic foundation decides:
- What concepts exist.
- What they mean.
- How they relate.
- What is primary and what is derived.
- Who gets to define them.
Most organisations assume this layer exists because the same words appear everywhere.
Using the same word is not the same thing as sharing the same definition.
You can use the same word in every meeting and still be having five different conversations.
Parallel Truth begins there.
It rarely begins with conflict.
It begins with convenience.
And convenience, left ungoverned, becomes architecture.
Every shortcut is a design decision in disguise. Some of them just take longer to admit it.
How Ambiguity Congeals
A concept is not fully agreed. One team implements it locally. Another encodes it slightly differently. A workaround becomes permanent because “we’ll rationalise it later”.
The domain model absorbs the compromise.
The system rarely forgets a compromise. It just promotes it.
Workflow is folded into status. Responsibility collapses into a single field. Derived values are stored as facts.
Nothing appears broken.
It simply feels efficient.
Then rules fork.
Permissions are re-implemented elsewhere. Calculations are copied. Transitions behave differently because this process is “special”… aren’t they all?
Now the system contains multiple formal interpretations of the same idea.
Parallel Truth is no longer a misunderstanding.
It is institutionalised.
And institutionalised ambiguity does not stay quiet.
The Propagation Of Parallel Truth
Atomic Systems: Elegant Amplifiers
Atomic content is powerful.
It allows meaning to be decomposed and recomposed across dashboards, notifications, APIs, reports, and increasingly, AI-generated surfaces.
Atomic systems assume stable meaning.
If a concept has one authority, reuse scales clarity.
If it has parallel definitions, reuse scales divergence.
Ambiguity becomes portable.
The more composable your architecture becomes, the more expensive Parallel Truth becomes.
Atomic design does not eliminate semantic instability.
It multiplies it.
The Design System Performs Certainty
Eventually the design system steps in.
- Badges align.
- Buttons align.
- Tables align.
- States look coherent.
From a distance, everything appears under control, but components cannot reconcile split authority. They can only render it consistently.
You can have the same badge meaning different things. The same action triggering different consequences. The same label implying different governance.
Design can make instability look disciplined, which is impressive but occasionally dangerous as it cannot make it disappear.
This is where organisations confuse visual order with conceptual agreement.
They are not the same thing.
What Parallel Truth Feels Like
Users do not experience domain models.
They experience predictability.
When the same word means different things in different places, something feels unreliable. When the same action behaves differently depending on context, trust erodes quietly.
No one reports “Parallel Truth Detected”.
They hesitate. They double-check. They create parallel processes.
Parallel Truth rarely produces drama.
It produces meetings.
And drift.
And drift, at scale, is governance failure expressed through interface.
Parallel Truth is not a UX inconvenience.
It is the system signalling that no one owns meaning.
A Sharper Diagnostic
Choose one shared concept, status is usually revealing.
Ask:
Where is it defined?
Who has authority over its meaning?
Where is it stored?
Where is it calculated?
Where is it enforced?
If the answers point to multiple owners, you do not have flexibility.
You have parallel authority.
A coherent system has one authority per concept.
Everything else is interpretation.
Two Architectures
Why This Matters Now
AI can generate interfaces. It can generate structured content. It can generate entire surfaces faster than most organisations can reason about them.
It cannot resolve Parallel Truth.
In fact, it operates comfortably inside it.
Large models are adept at producing plausible interpretations of ambiguous concepts. If your semantic layer is unstable, automation multiplies divergence across every output channel.
The more automated your surfaces become, the less tolerance your system has for split authority.
Design systems are becoming easier to produce.
Semantic discipline is becoming more decisive.
And more rare.
A Different Definition Of Consistency
Consistency is not visual similarity.
It is the absence of Parallel Truth.
When meaning has one authority: The model stabilises. Rules centralise. Atomic systems scale clarity. The design system expresses something real. The interface behaves predictably.
Coherence propagates but so does instability.
The difference is whether meaning has one home - or several.
Reflection
If your design system vanished tomorrow, what would remain stable?
A single, shared definition of what things mean?
Or a collection of aligned components sitting on parallel truths?
You can standardise presentation indefinitely.
If meaning has more than one authority, you are not consistent.
You are merely aligned.