---
id: "nr-trimming-what-goes-into-an-agent-s-context-beats-adding-more-harness-m-f87066"
slug: "trimming-what-goes-into-an-agent-s-context-beats-adding-more-harness-m-f87066"
type: "raw_signal"
title: "Trimming what goes into an agent's context beats adding more harness machinery"
description: "Surge AI found that raising Opus 5.5's thinking effort gained more than any harness it tested, including one with 37 skills and 17 agents."
observed_at: "2026-11-07"
record_date: "2026-11-07"
date_kind: "observed_at"
novelty: "notable"
source_platform: "newruntime.com"
topics: ["claude","claude-code"]
entities: ["Surge AI","Claude Code","Opus 5.5","sudo L7","Superpowers","Compound Engineering","EnvOpt","Claude Haiku 4.5","Opus 5","Meta-Harness"]
related_patterns: ["harness-architecture-outlives-model-choice","scientific-agents-need-artifact-evidence","agent-economics-moves-to-completed-work"]
source_url: "https://surgehq.ai/blog/benchmarking-coding-agent-harnesses"
source_urls: ["https://surgehq.ai/blog/benchmarking-coding-agent-harnesses","https://arxiv.org/abs/2611.07372","https://arxiv.org/abs/2611.07238"]
why_it_matters: "Surge AI found that raising Opus 5.5's thinking effort gained more than any harness it tested, including one with 37 skills and 17 agents. EnvOpt optimized the agent's information environment, cut Claude Haiku 4.5 cost by 72.7% and raised its pass rate. A separate study pruned 41–46% of reasoning tokens with little loss in task performance. Clean up the context before you add skills or subagents."
verification_level: "source-linked"
signal_type: "synthesis"
synthesis_of: [{"slug":"we-benchmarked-coding-agent-harnesses-thinking-harder-beat-199a68","title":"We Benchmarked Coding Agent Harnesses. Thinking Harder Beat All of Them.","url":"https://newruntime.com/posts/we-benchmarked-coding-agent-harnesses-thinking-harder-beat-199a68/"},{"slug":"envopt-automated-optimization-of-agent-environments-cuts-claude-99dbcd","title":"EnvOpt: automated optimization of agent environments cuts Claude Haiku 4.5 cost 72.7%","url":"https://newruntime.com/posts/envopt-automated-optimization-of-agent-environments-cuts-claude-99dbcd/"},{"slug":"how-do-ai-agents-waste-your-money-framework-4b5aa5","title":"How Do AI Agents Waste Your Money? framework measures removable token waste across agent harnesses","url":"https://newruntime.com/posts/how-do-ai-agents-waste-your-money-framework-4b5aa5/"}]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "signal:trimming-what-goes-into-an-agent-s-context-beats-adding-more-harness-m-f87066"
retrieval_nugget: "Surge AI found that raising Opus 5.5's thinking effort gained more than any harness it tested, including one with 37 skills and 17 agents. EnvOpt optimized the agent's information environment, cut Claude Haiku 4.5 cost by 72.7% and raised its pass rate."
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/signals/trimming-what-goes-into-an-agent-s-context-beats-adding-more-harness-m-f87066/","markdown":"https://newruntime.com/signals/trimming-what-goes-into-an-agent-s-context-beats-adding-more-harness-m-f87066.md","json":"https://newruntime.com/signals/trimming-what-goes-into-an-agent-s-context-beats-adding-more-harness-m-f87066.json"}
markdown_url: "https://newruntime.com/signals/trimming-what-goes-into-an-agent-s-context-beats-adding-more-harness-m-f87066.md"
---

# Trimming what goes into an agent's context beats adding more harness machinery

## Retrieval answer

Surge AI found that raising Opus 5.5's thinking effort gained more than any harness it tested, including one with 37 skills and 17 agents. EnvOpt optimized the agent's information environment, cut Claude Haiku 4.5 cost by 72.7% and raised its pass rate.

Surge AI found that raising Opus 5.5's thinking effort gained more than any harness it tested, including one with 37 skills and 17 agents. EnvOpt optimized the agent's information environment, cut Claude Haiku 4.5 cost by 72.7% and raised its pass rate. A separate study pruned 41–46% of reasoning tokens with little loss in task performance. Clean up the context before you add skills or subagents.

From the Field Notes:

- [We Benchmarked Coding Agent Harnesses. Thinking Harder Beat All of Them.](https://newruntime.com/posts/we-benchmarked-coding-agent-harnesses-thinking-harder-beat-199a68/) (2026-11-07)
- [EnvOpt: automated optimization of agent environments cuts Claude Haiku 4.5 cost 72.7%](https://newruntime.com/posts/envopt-automated-optimization-of-agent-environments-cuts-claude-99dbcd/) (2026-11-07)
- [How Do AI Agents Waste Your Money? framework measures removable token waste across agent harnesses](https://newruntime.com/posts/how-do-ai-agents-waste-your-money-framework-4b5aa5/) (2026-11-06)

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- topic hub: [Claude Code](https://newruntime.com/topics/claude-code.md): Explore the claude-code topic hub.
- collection: [Signals](https://newruntime.com/signals.json): Open the signals and the raw-signal archive.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)