Glossary · Plain definitions

Words, defined without the fog.

Terms used across this site. Where a term describes something we do not have, the definition says so.

Agent: Software that can take actions toward goals, not just answer questions.
Agent context: The memory, strategy, constraints and authority an agent operates with.
Scoped authority: Delegated permission bounded by action, amount, asset, time and revocation.
Local-first memory: Agent memory stored encrypted on the user's device, exportable and deletable.
Private intent: Keeping a planned action hidden until it executes.
Selective disclosure: Releasing scoped facts to an auditor without the full context.
Settlement: The public, on-chain result of an executed action.
Robinhood Chain: The network this project builds on. We are independent, not affiliated.
PONS: The launch platform configured for this project's token. The token is not launched.
USDG: The quote asset configured for this project. A fact, not a product claim.
CA: Contract address of a token. Not launched; nothing to verify yet.
TEE: Trusted execution environment — hardware that keeps computation hidden even from its operator. We make no TEE claim.
Enclave: An instance of confidential computing. Referenced as research, not as something we run.
Attestation: A signed report proving code ran in an enclave. We publish none.
Commit-reveal: Publishing a hash of an intent first, revealing it after execution.
Threshold encryption: Ciphertext openable only when a quorum of key holders agrees.
Zero-knowledge proof: Proof that a statement is true without revealing why it is true.
Threat model: A structured account of what you protect and from whom.
Trust boundary: The line across which you must trust something you do not control.
Research phase: This project's current stage: no product, no guarantees, published questions.