Articles

Notes on the certified semantic contract.

Short pieces on how Terazu works and why, most with screens from the real product: trust, definitions, your estate, security, and how to start.

Getting started

Who Terazu is for

The estate, the moments people go looking, the people inside who use it, and who it isn’t for yet.

Read
Getting started

Start with the numbers you already argue about

Why a pilot starts with the numbers people argue about, reads the whole estate within a read budget, and never needs a domain first.

Read
How it works

Earned context

Ingest, chase, certify, serve, decay: how a question pulls meaning into existence, and why nothing counts as certified until a person signs.

Read
How it works

The next best step

Home’s next acts and the open doors: what to define and capture next, what it unlocks, and what it will cost.

Read
How it works

Inside a chase

What happens when the certified pieces can’t work a question out, step by step, with a real chase from our earlier demo estate.

Read
Trust

The refusal is the feature

Why an AI that says “not yet” when it can’t pin a number down is worth more than one that always has a number.

Read
Trust

Anatomy of a receipt

What each part of an answer’s receipt tells you (how it was read, the pin, the source, who certified it), and what it doesn’t.

Read
Trust

When two numbers disagree

The same number in two places drifts. Terazu measures the gap where each one lives, for one number or every number in the workspace, and leaves the call to the people who own it.

Read
Definitions

Sorting the pile

Why Terazu won’t certify for you, and how it sorts, measures and prepares the pile so people decide only what needs a person.

Read
Definitions

Define a number by its shape

Pick a shape, point it at a meaning, and certify the sentence, not the code.

Read
Definitions

Your company’s words

LFL sales, card holder, till transaction: how Terazu learns your words, and why only certified ones count.

Read
Definitions

One meaning, and everything built on it

Open one meaning to see every table and number that depends on it, and follow it back to where its data is born.

Read
Definitions

Let the questions decide

Coverage shows which questions answer, which don’t, and what would unblock the most.

Read
Your estate

What Terazu can see, and what it can’t

The Estate Atlas: every system, the evidence behind each line between them, and the blind spots, named.

Read
Your estate

Bring your Power BI model

Your model already declares meanings and joins. Terazu reads it as evidence, rebuilds the DAX it recognizes, and certifies nothing on import.

Read
Security

Who may ask what

Five roles, sign-in through your identity provider, an audience on every definition, certification for people only, and AI tools outside Terazu that see shape, not content.

Read
Security

Keep your data away from the model

Metadata only, for the whole workspace or source by source; Restricted and flagged columns kept from models that aren’t local; AI tools outside Terazu see shape, not content. People keep their answers.

Read

Start a pilot

See the ideas on your estate.

The numbers you already argue about, the systems behind them, and your people certifying. Terazu runs on your servers, read-only against your data.