Cryptographic audit receipts for AI coding agents. Ed25519 + Merkle + RFC 3161 TSA. Supports Claude Code, Codex CLI & Cursor.

Append-only event kernel with Ed25519-signed Merkle checkpoints. Every AI action gets a verifiable receipt.

ATLAST Protocol — The Trust Layer for the Agent Economy. Make AI agent work verifiable with Evidence Chain Protocol (ECP). Open source · MIT License · weba0.com

Local autonomous AI. 15k+ Python lines. Just values. Proactive Shell/CDP agency. NO FRAMEWORK/CRONJOB/BEHAVIORAL PROMPT. Autonomy grows where allowed.

AI made a mistake. Can you prove what went wrong, and who is responsible? An open proposal for AI accountability management — lifecycle records, responsibility boundaries, tamper-evident audit trail. Not a finished answer.

Stdlib-only Python contracts for evidence packets, receipts, gates, ledgers, delegation chains, and witness receipts. Eleven domain wedges turn evidence your tools already produce into re-derivable proof packets: MATCH, DRIFT, or UNVERIFIABLE. Reached through the telos proof CLI.

Gate AI-agent actions with explicit grants, durable journals, MCP server support, and verification.

Accountability layer for multi-agent AI systems. Content-addressed, cryptographically signed claim DAGs.

Turn local files, git state, and Markdown claims into hash-stamped provenance receipts.

Cryptographic receipt system for AI agent accountability. Tamper-evident, hash-chained receipts with Ed25519/HMAC signing.

Audit AI-agent sessions by comparing declared intent, action ledgers, and scope policy.

A public doctrine for treatment norms in machine-logic systems under moral uncertainty.

Generate and check GitHub CODEOWNERS from accountable surface declarations.

AI governance framework built on three architectural principles: Commander's Intent, Observable Autonomy, and Convergence Is Silence. Defines Autonomous Reasoning Fidelity (ARF) as the target property.

Generate replayable creative worlds with shaders, sound, motion timelines, and receipts.

Turn your own material into a runnable course: FSRS spaced repetition, retrieval practice, real grading, zero dependencies. Halts hard at every graded step, so the machine never takes the test for you; every graded step leaves a re-verifiable receipt.

Sutton audio and power llc

LUMINA-30: non-binding boundary framework for preserving effective human refusal before irreversible AI consequences.

Validate model and release claims against small provenance envelopes with redacted output.

∈ Principle — A philosophical and institutional proposal for public authorship and responsibility in the age of AI. Foundational text with DOI (Zenodo). CC BY 4.0.