cybersealnull/chord-affect-anchors

Chord Affect Anchors is a research prototype exploring whether musical chord notation can serve as a compact, portable language for describing emotional states in AI conversations. The core problem it addresses is that large language model (LLM) agents lose emotional context between sessions: if you talk to an AI today about something that felt bittersweet, and then start a new conversation tomorrow, the model has no memory of that emotional texture. The proposed solution is a two-line notation unit: one concrete context sentence describing a situation, followed by one line of musical chord symbols (like "Fmaj9 to C/E to Am add9, 60bpm"). The claim is that major language models share learned associations between musical chords and emotional moods from their training data, so a chord progression can act as a shared emotional shorthand that different models from different providers can interpret in roughly the same way, without needing any external tool, embedding model, or database. The approach is entirely text-native: chord notation is plain ASCII that fits inside any document or prompt. The README notes a small anecdotal pilot across six AI readers from five providers (Anthropic, OpenAI, ByteDance, DeepSeek, Google), finding that adding more context alongside the chord made the emotional interpretation more precise. The repository contains a long-form article in English and Chinese, an X (Twitter) thread draft, and an HTML slide deck. It is a conceptual and writing project rather than a code library. No installation is required.