Agents2026-06-20 · source-backed

For tool-heavy agents, a recency window plus a running summary beats a retriever, with no external store.

Issue

2026-06-20

Confidence

source-backed

Sources

Redaction

redacted

Story

"Less Context, Better Agents" shows that long-horizon, tool-using agents drowning in verbose tool responses complete tasks reliably with just a recency window and a compact running summary (arXiv:2606.10209). This is a useful corrective to the reflex of bolting RAG onto everything. Before you stand up a vector store for your agent's working memory, try recency plus summary and measure whether you actually lose anything. Most of the time you won't, and you'll have skipped a whole moving part.

Source trail

arXiv:2606.10209

Entities

Most RAG

Provenance

Canonical issue: Ramsay Research Agent — June 20, 2026
AI generated: no
Story unit: 2026-06-20-for-tool-heavy-agents-a-recency-window-plus-a-running-summary-beats-a-retriever-with-no-external
Labels: source-backed, canonical briefing excerpt

For tool-heavy agents, a recency window plus a running summary beats a retriever, with no external store.

Story

Related stories

MetaSyn shows agents recover under 53% of relevant literature despite 91% retrieval recall.

10 actionable skills from today's findings:

Evaluate causal-attribution guardrails for RAG and tool agents.

Source trail

Entities

Provenance

Red-team your RAG with deliberately poisoned retrieval.

Microsoft SkillOpt Trains Markdown Skills for Frozen LLMs. 52/52 Benchmark Wins. Your CLAUDE.md Can Be Optimized Like a Neural Network.

Knowledge Objects: Hash-Addressed Facts Hit 100% Accuracy at 252x Lower Cost Than In-Context

Engram: a lean retrieved context beats stuffing the full history into the window.

Don't bolt agentic enhancements onto RAG uniformly.