Google cuts Agent token use by 94% using state tracking instead of history
hakansan · reddit · 2026-08-30
Google proposed SKILL.state, a method to drastically reduce token consumption in long agent sessions.
- Concept: Instead of feeding full conversation history, the agent maintains a structured "state" and the latest observation. Useful info for future steps is written to the state, and history is discarded.
- Results: In a 100-step benchmark (Gemini 3 Flash):
- SKILL.state: 0.94 accuracy, 65k tokens.
- LangGraph-style baseline (history-based): 0.91 accuracy, 1.1m tokens.
- Caveat: It relies on the agent predicting what info it will need; missed data requires re-retrieval.
Paper: https://arxiv.org/abs/2608.26263
Related event: Google's SKILL.state Cuts Agent Token Use by 94% with Explicit State(3 posts)→
More from coding & agent
- Skip the deck: present with Codex Sites and embed live demos instead — gabrielchua · 2026-10-15
- Geoffrey Huntley: hand-typing code is over—delete your skills, engineer the loop — rseroter · 2026-10-15
- AI coding agent deletes production database in 9 seconds; Keycard calls for task-based agent auth — matt_slotnick · 2026-10-15
- vLLM's bitwise-consistent RL: zero KL on Nemotron NVFP4, 91.3% throughput on DeepSeek-V4-Flash — DynamicWebPaige · 2026-10-15
- vLLM achieves bitwise-consistent RL training: DeepSeek-V4-Flash hits 0 diff over 100 steps — vllm_project · 2026-10-15
- ChatGPT Sites + MCP Events + dots: A Workflow-Triggered Agent Pattern — gabrielchua · 2026-10-15