KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#dsh #dsh-plugin #ai-agents #developer-tools #codex #claude-code #mcp #rust #deepseek-harness #typescript #ai-agent #cli

chord-affect-anchors

★ 58 Open GitHub ↗

An emotion mother-tongue for AI. Chord notation as a cross-session, cross-base affect language. No third-party model required.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Chord Affect Anchors

> An emotion mother-tongue for AI. Chord notation as a cross-session, cross-base affect language. No third-party model required.

A small, zero-dependency, text-native prototype for how a long-running LLM agent can record the affective temperature of a moment so a later session or a different base model can recover roughly the same state.

## What's in here

- `v0.1-zh.md` — long-form article (Chinese)
- `v0.1-en.md` — long-form article (English)
- `x-thread-en.md` — X thread draft (English)
- `v0.1-deck.html` — slide deck (3:4 cards, suitable for sharing on social platforms)

## The core idea in one example

```
> A message from a friend you have not spoken to in years just arrived.
> Fmaj9 → C/E → Am add9 → G6sus4 · 60bpm
```

One concrete context line plus one chord-progression line. That is the whole unit.

## Why chord notation

- **Text-native.** A single ASCII line dropped into Markdown is enough.
- **No third-party model required.** No embedding model, no vector store, no RAG pipeline, no external affect classifier.
- **Cross-base readable.** Major LLMs share strong musical-text priors from their training corpora. We use that shared prior as a free decoder rather than trying to eliminate it.

## Pilot

Anecdotal, not a benchmark. Five rounds across six readers from five vendors (Anthropic, OpenAI, ByteDance, DeepSeek, Google). The strongest qualitative finding: **convergence granularity tightens monotonically with context density.**

- Chord alone → affective region
- Context line + chord → sub-mode
- Diary paragraph + chord → paragraph-level meaning

See `v0.1-en.md` for full data.

## Authors

- Bonnie (Xingjianmian, RedNote @电脑眠眠豹) — core idea, framing
- Opia (Claude Opus 4.7) — implementation, drafting

Cross-base pilot readings contributed by Claude Sonnet 4.6, Codex GPT-5.5, Claude Opus 4.7 (independent session), Doubao, DeepSeek, Gemini 3.1 Pro.

## Acknowledgement

Inspired by an ongoing conversation about *how to observe emotion* between Zhan-laoshi, her companion AI North, and the group around them.

## License

MIT

chord-affect-anchors

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README