Research
Harmonic Coordination Theory (HCT) proposes a musical ontology for multi-agent coordination. Current frameworks give you orchestration tools but lack a shared language for timing, quality, and intent.
I fix this using cues, fermatas, tempo, and dissonance as engineering primitives.
Key Publications
- Harmonic Coordination Theory — CS.AI / CS.MA papers
- stefanwiest/hct-spec — Canonical signal specification
- stefanwiest/hct-core — Reference implementation
Protocol Extensions
- stefanwiest/hct-mcp-signals — Coordination layer for Anthropic MCP
- stefanwiest/hct-a2a — Coordination semantics for Google A2A
Tools
Production tools that solve real problems—usually ones I've encountered myself.
| Tool | Purpose | Link |
|---|---|---|
| AntiSlop | Multi-language linter for AI-generated code slop | GitHub |
| EnvCheck | Environment file linting for DevSecOps | GitHub |
| NerfStatus | Scientific LLM degradation detection | Web |
| SpeakOps | Real-time AI speech coaching | GitHub |
Writing
SeekingSota — Weekly essays on AI × Human identity crisis
I explore what happens when engineers stop programming and start conducting AI agents. Real agent workflows, technical deep dives, philosophy of production, and honest stories about the identity crisis that comes with this transformation.
HuggingFace
Models & Datasets (stefanwiest/)
- stefanwiest/hct-spec — HCT specification & signal reference
Spaces (skew202/)
- skew202/nerfstatus-hf-monitor — HF inference quality monitor
Collections
Curated research collections grounding multi-agent systems in academic literature:
- Multi-Agent Coordination & Signaling — A2A, MCP, CAMEL, IoA, HCT protocols
- LLM Reasoning & Planning Techniques — CoT, ToT, GoT, ReAct, DSPy
- LLM Quality & Degradation Monitoring — Hallucination detection, model collapse, probe-based testing
- Memory, Context & RAG — RAG architectures, MemGPT, semantic chunking
- Tool Use & Agent Execution — Toolformer, MCP, agentic loops
- Agent Safety & Alignment — Jailbreak detection, guardrails, constitutional AI