TokenWire
A research note on moving evidence, not noise, between retrieval, planning, and generation layers.
Token Efficiency RAG Systems Compression
[RESEARCH]
RESEARCH MODE: SYSTEMS TO INTELLIGENCE
研究Thin-grid notes on efficient tokens, small models, agent runtimes, RAG evaluation, voice AI, and architecture experiments.
Small Language Model Experiments
Exploring where small models can reduce latency and cost without pretending to be general intelligence.
SLM Routing Evaluation Cost
Agent Runtime Notes
Lab notes for making agent loops observable and recoverable.
Agents Durability Tools Runtime
RAG Evaluation Notes
A research notebook on retrieval quality, provenance, and answer evaluation.
RAG Evaluation Search Grounding
Voice AI Systems
Notes on voice-first AI systems where timing is part of correctness.
Voice AI Realtime UX Agents