---
title: "praedicamenta — Manifesto | praedicamenta"
canonical_url: "https://praedicamenta.com/manifesto"
last_updated: "2026-08-05T18:07:50.499Z"
meta:
  description: "Knowledge is scattered and always will be. Context is not the place to fix it. Fix it at the right layer."
  "og:description": "Knowledge is scattered and always will be. Fix it at the right layer."
  "og:title": "praedicamenta — Manifesto"
---

praedicamenta — home

praedicamenta on LinkedIn

Manifesto

# Knowledge is always scattered**.**<br>Context doesn't fix it**.**<br>Fix it at the right layer**.**

> *“If HP knew what HP knows, we would be three times as profitable” *

He said it decades ago. The insight hasn't aged: your company doesn't necessarily know what it knows. If it did, it wouldn't just answer faster — it would decide better. That knowledge is the asset.

§Principles

01

### Knowledge is scattered.<br>Always will be.

There is no single source of knowledge, no single system of record for a business, and there never will be: every application, every team, every warehouse, even the unwritten, holds its own slice of what the company knows.

Knowledge is also unbounded and ever-changing. Ambiguity isn't something to fix — it's something to detect and interpret. What was true yesterday and what's true today can both be real at once, and both have to be preserved, not overwritten. That makes fragmentation an engineering problem, not a one-time project: something to encapsulate in the right abstraction, not solve and forget.

02

### Fix it at the right layer.<br>

Bolting context onto each AI tool separately doesn't solve fragmentation — it just relocates it, tool by tool.

There is a gap between intelligence and storage that none is filling. We expect our intelligent systems to remember everything, but their context is too small. Our storage systems are practically unbounded, but unequipped to help us reason about what they hold. Something has to sit between the two.

That something is a shared abstraction — knowledge shouldn't be caged inside any one application. It should be common ground: something humans, agents, and applications all draw from directly. Every new tool, every new teammate, every new agent would otherwise add complexity quadratically. Unify once against that shared layer instead, and the cost stays linear — paid once, reused everywhere.

03

### Enterprise knowledge is a garden.<br>Context is what blooms from it.

Knowledge is the asset — it's what grows and compounds over time, weaving connections underneath the soil, just waiting to be activated.

Context is momentary. It's a slice of that knowledge, doing its job when it's activated at the right time, for the right task, in the right "<< context >>".

Context blooms from the rich soil that is knowledge, and it doesn't need to be regrown from a different soil every time.

04

### The model is a commodity.<br>Knowledge compounds like an asset — or evaporates.

Every competitor can rent the same model you can. What isn't for sale is what your company knows — the history, the exceptions, the judgment calls sitting in people's heads and in a decade of systems. You don't need to make the same mistake twice, or come to the same conclusion twice — you move faster.

Insight that stays buried in a chat thread or a single person's head is knowledge the company already paid for and still doesn't have. What one person figures out becomes infrastructure the whole team, and every future agent, can start from. One person's win becomes the whole team's — or it gets lost, and the team has to reach the same conclusion over and over again.

05

### Knowledge is compression.<br>Compression is an iceberg.

Knowledge is a compressed expression of expertise — years of judgment, failed attempts and exceptions, folded into something that can be packaged, distributed, and executed directly. It's the new unit of economics.

The visible surface is small, the value is deep.

Compression is the work, the goal, and the asset.

06

### Access is not meaning.<br>

Reach isn't understanding. APIs, connectors, and agent frameworks make giving an AI access to every system the easy part — that's commodity work by now. What breaks is meaning: an agent can touch all of it and still not know what any of it means. Understanding takes more than storing and retrieving information; it takes knowing how the pieces relate, why they matter, and what changed since the last time anyone looked. Models aren't magic — to decide well, they need that understanding already assembled somewhere, not reconstructed one query at a time.

## **This is what we are building. **

A unified enterprise workspace — one shared ground where storage, search, and understanding converge so that humans and agents can reason over them easily. Unify once, reuse everywhere.

[See the solution→](https://praedicamenta.com/#solution)