Voice & Terminology

Welcome to the Inference Voice Guidelines

How we write: headlines, claims, figures and the words we use.

</aside>

Voice

Verifiable, precise, principled, minimal, non-corporate. Say less; prove more.

Headlines are claims. Full sentences with a period, one per slide or section: "Every technical milestone hit." "Code generation is largely solved. Acceptance is the bottleneck."
Claim, then receipt. A figure sits beside what it counts: "TruthTensor put 1.1M agents on our rails." Every number must trace to a document.
Concrete over vague. "Cryptographically verifiable", not "secure". "Proven", not "trusted".
No buzzword stacks. Never "leverage", "synergy", "ecosystem play", "revolutionary".
Machines don't think. Not "the AI thinks" or "the model wants".
Active voice. Say who did what.
Sentence case for headings and buttons. No emoji in product, decks or memos.

Term	Use
Inference Labs	The company, in prose and legal copy. The wordmark reads "inference"
Proof of Inference	Title Case, a proper noun
Auditable Autonomy	Title Case, the company thesis
Inference Network	Title Case, the network product
Sertn	The verifiable computer vision product
DSperse	Exact casing. Targeted verification for zkML
zkML	Lowercase z
veiled model weights	Lowercase
`>>> inference labs`	CLI signature, lowercase, in mono