a Parable project

Every platform models
storage.
None model thinking.

Perceptions are a data model for the human mind — where Concepts, Relationships, and time are first-class, and every platform on earth is an implementation detail.

People, Work, Goals & Relationships — not tables, columns & joins.
Snapshot, activity & comparison — three kinds of time, not one.
Identity resolved. Types enforced. Every answer traceable to source.
Knowledge in minds, meetings & documents — captured and aligned.
Experience Perceptions
Scroll
The foundation

Built for how
humans think

Almost every question an executive asks involves relationships and time. "Who on my team sent the most messages last quarter?" — that's a relationship (who is on my team), a traversal (sent), and a temporal filter (last quarter). No platform has ever been designed for this. So everyone flattens everything into tables and builds workarounds to get the structure back.

Perceptions model your organization the way it actually works. People, Work, Goals, Actions — hierarchical, strongly typed, with identity resolution across every tool. You traverse relationships, not joins. Every attribute typed with semantic meaning (Email, Dollar, URL — not varchar). Temporal modes are explicit: snapshot, activity, and comparison.

The problem

Context graphs and
semantic layers failed

Context graphs inherit 25 years of knowledge graph failure modes. Enterprise knowledge graph projects fail at an estimated 80% rate. The Semantic Web never achieved enterprise adoption despite billions in investment. Context graphs are non-deterministic — the same traversal produces different results — and they're still built from the storage model, not the human model.

Semantic layers (LookML, dbt, Cube.js) add flexibility in the wrong places: no opinion, unlimited views on top of data. Entire companies spin up to manage the views of other companies. And none of them span action — a context graph can tell you a column contains PII but can't mask it. A semantic layer can define a metric but can't send an alert when it changes.

Every other approach
→ Context graphs: 80% failure rate, non-deterministic, storage-model-first
→ Semantic layers: another layer of SQL, no opinion, no action
→ Query builders: thin paint on developer tools, modality cliff
Perceptions
A data model for the human mind. Not the storage model.
↳ Opinionated concepts. Traversal not joins. Every platform is an implementation detail.
How it works

Structure first.
Your expertise on top.

P
Parable
Done for You
People, Work, Goals — the graph, modeled. Strongly typed (Email, Dollar, URL). Identity resolution built in. Temporal modes explicit. This is the infrastructure you'd spend a year building. It's already here.
Y
Your Classifications
Your Domain
Layer your domain expertise as typed, structured classifications. Capture context — upload documents, paste instructions, brain dump what's in your head. The system proposes categories. You review, preview on a sample, and approve. Not SQL views — structured knowledge with provenance.
▲
Plots
Extensible Contracts
Every transformation is a YAML contract. Every function is opinionated and extensible — one send() for email, SMS, Slack. One classify() for any concept. The contract absorbs mistakes: if we designed the wrong function, that's a new extension, not a new architecture.
The difference
The platform never dictates the architecture.
Postgres, Iceberg, Redis, Kafka — these are implementation details. Swap any of them without changing how you think about your organization. Perceptions sit above every platform. Concepts are stable. Vendor schemas change underneath. The human model is the architecture.
Full lineage
Full provenance. Every step.
Your output → Classification → Parable Graph → Provider Plot → Raw source data
The architecture

How you get
your time back

01

Concepts, not tables

You work with People, Work, Goals, and Relationships — not salesforce_contacts and jira_issues. When Salesforce renames a field, you update one Provider configuration. Your analysis doesn't know or care. The human model is stable. Vendor schemas are volatile. Build against the stable thing.

02

Traversal, not joins

"Who on my team sent the most messages last quarter?" is a traversal: follow the relationship, filter by time. Not a three-table JOIN with a GROUP BY. Relationships are first-class in the graph. You ask questions the way you think about them. The system handles the rest.

03

Three kinds of time, not one

Snapshot: what did the data look like then? Activity: what happened during that period? Comparison: how does then compare to now? Most systems pick one and pretend the others don't exist. Temporal behavior is a property of the Concept. The contract makes the semantics explicit.

04

Knowledge lives everywhere

Not all knowledge is in the data. Some is in documents. Some is in people's heads. Some disagrees across source systems. Classify captures all three: structured data, uploaded artifacts, and unstructured brain dumps. The Perception grows structurally — not just in metadata, but in the graph itself.

The action layer

Perceptions define.
Plots act.

Every platform on Earth models storage. Perceptions model thinking. Concepts are stable. Platforms are implementation details. Identity resolved. Types enforced. Temporal modes explicit. Build against how humans think, not how data is stored.

Explore Plots
◯
Perceptions
Define
▲
Plots
Act

Structure first.
Analysis on top.