---
title: "Aftermarket Harnesses"
description: "The harness, not the model, is now the faster-moving variable in AI cost \u0026 capability. It owns the prefix, \u0026 the prefix is 86-98% of the bill."
categories: ["AI"]
keywords: ["AI harness","Claude Code","Cursor","Codex","prompt caching","input tokens","OpenRouter","Endor Labs","Agent Security League","GPT-5.5","Claude Opus 4.7"]
ai_summary: "The harness now moves the coding benchmark more than the model does. Endor Labs' Agent Security League found GPT-5.5 scored 61.5% functional correctness in Codex \u0026 87.2% in Cursor, \u0026 Claude Opus 4.7 scored 87.2% in Claude Code \u0026 91.1% in Cursor. Input tokens are 86-98% of OpenRouter volume, so the harness controls most of the bill through cache discipline. First-party co-design buys real cache hit rates, but a third-party harness can match them."
date: 2026-07-28
lastmod: 2026-07-28
canonical_url: https://www.tomtunguz.com/aftermarket-harnesses/
author: "Tomasz Tunguz"
---


AI harnesses have more impact on performance than the models.

Endor Labs ran the same models through two harnesses in the same week. OpenAI's GPT-5.5 scored 61.5% functional correctness in its native Codex harness & 87.2% in Cursor's, a 25.7-point swing from the runtime alone. Anthropic's Opus 4.7 scored 87.2% in Claude Code & 91.1% in Cursor.[^1]

Both frontier models performed better in a competitor's harness than in the one their maker ships.

Harnesses are a fulcrum in the AI stack. They influence cost, quality, & accuracy.

Input tokens are 86-98% of LLM traffic on OpenRouter. Output costs 5x input per token, but input still dominates the bill because there is so much more of it. One year ago I wondered about this ratio in the Hungry, Hungry AI Model.[^2] Practitioners said 95% input/5% output. At scale, they were right.

{{< email_image src="hrdleti0co668gwid4cg" alt="Stacked bar chart showing share of token volume by type for five OpenRouter models. Input tokens account for 85.7% to 98.2% of volume, with reasoning and completion making up the remainder." width="540" height="342" >}}

Controlling input costs is therefore a valuable proposition. The model doesn't control those costs, but the harness does.[^3] Harnesses decide which context to send. Much of that context repeats from query to query. Caching it intelligently can save 40-80%. A study across 500 long-horizon agent sessions found 41-80% cost reduction & 13-31% faster time-to-first-token, with savings scaling linearly from 500- to 50,000-token prompts.[^4] The winner was caching only the stable prefix & placing dynamic content after the cache breakpoint.

Harnesses also perform information retrieval: the code to read, the style guide, the section of an investment brief to analyze. The more concise & precise, the lower the cost.

Cursor's harness matches Claude Code technique-for-technique: dynamic tool fetching, priority-based prefix assembly, two-tier caching. That is why Opus 4.7 scored higher in Cursor than in Claude Code, & why GPT-5.5 nearly doubled its functional-correctness score outside Codex.

The bundled case is not without merit. Claude Code treats cache hit rate as an uptime metric, hits about 96% in real sessions, shares the system-prompt cache across users on the same version, & builds forked sub-agents at 99% byte-identity for 90% savings.[^5]

Co-designing the harness, cache API, & model buys real cache discipline. But the discipline lives in the harness, & a third-party harness can match it, as Cursor's numbers show.

The harness has shed its critique as a model wrapper ; it's a jockey pushing AI performance further than the breeder imagined.

[^1]: Endor Labs, "GPT-5.5 Sets a New Code Security Record with Cursor, not Codex, in Agent Security League," Apr 27, 2026, https://www.endorlabs.com/learn/gpt-5-5-sets-a-new-code-security-record-with-cursor-not-codex-in-agent-security-league. Agent Security League benchmark, based on Carnegie Mellon's Open SusVibes framework.

[^2]: Tomasz Tunguz, "The Hungry, Hungry AI Model," July 8, 2025, https://tomtunguz.com/input-output-ratio/

[^3]: Tomasz Tunguz, "The Harness Is the New Battleground," https://tomtunguz.com/the-harness-is-the-new-battleground/

[^4]: Lumer et al., "Don't Break the Cache: An Evaluation of Prompt Caching for Long-Horizon Agentic Tasks," arXiv:2601.06007, 2026. https://arxiv.org/abs/2601.06007

[^5]: Anthropic, "Lessons from building Claude Code: Prompt caching is everything," Apr 2026, https://claude.com/blog/lessons-from-building-claude-code-prompt-caching-is-everything; swap-tax figures from Digital Applied, "Prompt Caching Economics: Cache-First Agent Design," Jul 2026, https://www.digitalapplied.com/blog/prompt-caching-economics-cache-first-agent-architecture-2026.