---
schema_version: "newruntime-agent-readable-v0.2"
type: "raw_signal"
stable_id: "signal:openai-a-benchmark-score-reflects-the-model-as-well-as-the-harness-and-settings"
id: "x-2082616641834422740"
slug: "openai-a-benchmark-score-reflects-the-model-as-well-as-the-harness-and-settings"
title: "OpenAI: A benchmark score reflects the model as well as the harness and settings used to run it."
description: "A public X post from OpenAI with a linked primary source flags A benchmark score reflects the model as well as the harness and settings used to run it. For long-running agents, retaining reasoning and compacting context lets the model build o..."
retrieval_nugget: "A public X post from OpenAI with a linked primary source flags A benchmark score reflects the model as well as the harness and settings used to run it. For long-running agents, retaining reasoning and compacting context lets the model build o..."
observed_at: "2026-07-29"
record_date: "2026-07-29"
date_kind: "observed_at"
why_it_matters: "This X-discovered record adds fresh evidence to the agent ready software exposes capabilities, model answers become task interfaces lens and lets the site trend graph move as public product and research signals arrive."
novelty: "structural"
verification_level: "source-linked"
signal_type: "research"
evidence_kind: "creator-source"
status: "published"
source_platform: "x"
source_record_id: "2082616641834422740"
source_url: "https://x.com/OpenAI/status/2082616641834422740"
topics: ["generated-ui","evals","api-design","models","context-engineering","agents","interfaces"]
entities: ["OpenAI"]
related_patterns: ["agent-ready-software-exposes-capabilities","model-answers-become-task-interfaces","harness-architecture-outlives-model-choice","skills-become-portable-capability-layer","company-memory-needs-write-loops","verification-bandwidth-is-the-scarce-resource","agent-economics-moves-to-completed-work","goal-scoped-loops-replace-manual-continuation"]
source_urls: ["https://x.com/OpenAI/status/2082616641834422740","https://openai.com/index/how-two-settings-tripled-our-arc-agi-3-scores/"]
import_batch: "x-analyzer-20260730-084757"
routes: {"html":"https://newruntime.com/signals/openai-a-benchmark-score-reflects-the-model-as-well-as-the-harness-and-settings/","markdown":"https://newruntime.com/signals/openai-a-benchmark-score-reflects-the-model-as-well-as-the-harness-and-settings.md","json":"https://newruntime.com/signals/openai-a-benchmark-score-reflects-the-model-as-well-as-the-harness-and-settings.json"}
source_format: "x-api-normalized-json"
---

# OpenAI: A benchmark score reflects the model as well as the harness and settings used to run it.

## Retrieval answer

A public X post from OpenAI with a linked primary source flags A benchmark score reflects the model as well as the harness and settings used to run it. For long-running agents, retaining reasoning and compacting context lets the model build o...

## Observation

A public X post from OpenAI with a linked primary source flags A benchmark score reflects the model as well as the harness and settings used to run it. For long-running agents, retaining reasoning and compacting context lets the model build o...

## Why it matters

This X-discovered record adds fresh evidence to the agent ready software exposes capabilities, model answers become task interfaces lens and lets the site trend graph move as public product and research signals arrive.

## Provenance

This public record is an English normalization of an approved X watchlist post. Linked source_urls carry the publishable evidence boundary.