{
 "ledger": {
  "title": "Claim Ledger: Same Model, Different Product: The Case for Harness Engineering",
  "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/",
  "essay": {
   "title": "Same Model, Different Product: The Case for Harness Engineering",
   "url": "https://durabilitycurve.com/blog/harness-engineering-same-model-different-product/",
   "published": "2026-05-30"
  },
  "last_checked": "2026-10-03",
  "format": "current",
  "counts": {
   "claims": 14,
   "by_status": {
    "VERIFIED": 14
   },
   "verified_at_primary": 14,
   "removed": 0
  },
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "attribution": "The Durability Curve (Harry Floyd), https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/"
 },
 "rows": [
  {
   "id": "HE1",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE1",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "In the Meta-Harness paper's TerminalBench-2 table, Claude Opus 4.6 has pass rates of 58.0 under Claude Code (an official-leaderboard result the paper reports) and 76.4 under the Meta-Harness harness. Scope: Meta-Harness searched and was finally evaluated on the same 89-task TerminalBench-2 benchmark, and the paper describes the resulting harness as specialised to the TerminalBench-2 regime.",
   "quote": "search and final evaluation on the same 89-task benchmark … specialized to the TerminalBench-2 regime … Claude Opus 4.6 … Claude Code × 58.0 … Meta-Harness ✓ 76.4",
   "removal_reason": "",
   "source": "Lee, Nair, Zhang, Lee, Khattab and Finn, Meta-Harness: End-to-End Optimization of Model Harnesses, arXiv:2603.28052v1",
   "source_url": "https://arxiv.org/pdf/2603.28052",
   "locator": "p. 9, §4.3 Evaluating Agentic Coding Harnesses on TerminalBench-2 (setup paragraph) · Table 7, Claude Opus 4.6 block; caption: results of others are from the official leaderboard",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE2",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE2",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "In the same table, Claude Haiku 4.5 has pass rates of 27.5 under Claude Code (an official-leaderboard result the paper reports) and 37.6 under the Meta-Harness harness. Scope: Meta-Harness searched and was finally evaluated on the same 89-task TerminalBench-2 benchmark, and the paper describes the resulting harness as specialised to the TerminalBench-2 regime.",
   "quote": "search and final evaluation on the same 89-task benchmark … specialized to the TerminalBench-2 regime … Claude Haiku 4.5 … Claude Code × 27.5 … Meta-Harness ✓ 37.6",
   "removal_reason": "",
   "source": "Lee, Nair, Zhang, Lee, Khattab and Finn, Meta-Harness: End-to-End Optimization of Model Harnesses, arXiv:2603.28052v1",
   "source_url": "https://arxiv.org/pdf/2603.28052",
   "locator": "p. 9, §4.3 (setup paragraph) · Table 7, Claude Haiku 4.5 block; also Figure 1 (right), p. 1",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE3",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE3",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "LangChain reports improving its coding agent deepagents-cli by 13.7 points on Terminal Bench 2.0, from 52.8 to 66.5, by changing only the harness while keeping the model fixed at GPT-5.2-Codex.",
   "quote": "improve deepagents-cli … 13.7 points from 52.8 to 66.5 on Terminal Bench 2.0. We only tweaked the harness and kept the model fixed, gpt-5.2-codex",
   "removal_reason": "",
   "source": "LangChain (Vivek Trivedy), Improving Deep Agents with harness engineering, 17 February 2026",
   "source_url": "https://www.langchain.com/blog/improving-deep-agents-with-harness-engineering",
   "locator": "opening section, paragraph beginning \"We used a simple recipe\"",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE4",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE4",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "Claude Code's command reference documents /compact as a command that frees up context by summarising the conversation so far.",
   "quote": "/compact [instructions] … Free up context by summarizing the conversation so far.",
   "removal_reason": "",
   "source": "Claude Code Docs, Commands",
   "source_url": "https://code.claude.com/docs/en/commands",
   "locator": "command table, row \"/compact [instructions]\"",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE5",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE5",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "Claude Code's environment-variable reference documents CLAUDE_CODE_MAX_RETRIES, which overrides the number of times Claude Code retries failed API requests.",
   "quote": "CLAUDE_CODE_MAX_RETRIES … Override the number of times to retry failed API requests",
   "removal_reason": "",
   "source": "Claude Code Docs, Environment variables",
   "source_url": "https://code.claude.com/docs/en/env-vars",
   "locator": "Variables list, row \"CLAUDE_CODE_MAX_RETRIES\"",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE6",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE6",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "Claude Code's features overview describes subagents as isolated workers with their own context, running separately from the main conversation.",
   "quote": "Subagents … are isolated workers that run separately from your main conversation … Isolated worker with its own context",
   "removal_reason": "",
   "source": "Claude Code Docs, Extend Claude Code (features overview)",
   "source_url": "https://code.claude.com/docs/en/features-overview",
   "locator": "Compare similar features, tab \"Skill vs Subagent\", row \"What it is\"",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE7",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE7",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "Manus warns, from its own experience building an agent rather than from a measured experiment, that when users plug hundreds of tools into an agent's action space, the model is more likely to select the wrong action or take an inefficient path.",
   "quote": "hundreds of mysterious tools … the model is more likely to select the wrong action or take an inefficient path",
   "removal_reason": "",
   "source": "Yichao 'Peak' Ji, Manus, Context Engineering for AI Agents: Lessons from Building Manus, 18 July 2025",
   "source_url": "https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus",
   "locator": "section \"Mask, Don't Remove\", first paragraph",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE8",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE8",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "Manus deliberately leaves failed actions and errors in the agent's context, arguing that erasing failure removes the evidence the model needs to adapt.",
   "quote": "Erasing failure removes evidence. And without evidence, the model can't adapt. … leave the wrong turns in the context",
   "removal_reason": "",
   "source": "Yichao 'Peak' Ji, Manus, Context Engineering for AI Agents: Lessons from Building Manus, 18 July 2025",
   "source_url": "https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus",
   "locator": "section \"Keep the Wrong Stuff In\", second and third paragraphs",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE9",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE9",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "In Can Bölük's edit-format benchmark, the hashline edit format beat the patch format in 14 of 16 models.",
   "quote": "hashline beats patch in 14/16 models",
   "removal_reason": "",
   "source": "Can Bölük, Stencil, We improved 15 LLMs at coding in one afternoon. Only the harness changed., 12 February 2026",
   "source_url": "https://stencil.so/blog/the-harness-problem",
   "locator": "benchmark chart caption (3 runs × 180 tasks per model)",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE10",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE10",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "In the same benchmark, Grok Code Fast 1 went from 6.7% with the patch edit format to 68.3% with hashline.",
   "quote": "Grok Code Fast 1 went from 6.7% to 68.3%, a tenfold improvement … patch 6.7% … hashline 68.3%",
   "removal_reason": "",
   "source": "Can Bölük, Stencil, We improved 15 LLMs at coding in one afternoon. Only the harness changed., 12 February 2026",
   "source_url": "https://stencil.so/blog/the-harness-problem",
   "locator": "results paragraph beginning \"Sixteen models, three edit tools\" · chart row 01 point labels",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE11",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE11",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "Menlo Ventures estimated in December 2025 that Anthropic held 54% of the enterprise coding market, up from 42% six months earlier, driven in large part by Claude Code. The figure comes from the report's LLM market-share analysis, which estimates model-provider share from production API usage.",
   "quote": "coding market … an estimated 54% market share … up from 42% just six months ago, driven in large part by the popularity of Claude Code",
   "removal_reason": "",
   "source": "Menlo Ventures, 2025: The State of Generative AI in the Enterprise, 9 December 2025",
   "source_url": "https://menlovc.com/perspective/2025-the-state-of-generative-ai-in-the-enterprise/",
   "locator": "section \"LLM Market Share: Anthropic Extends Its Lead in the Enterprise\", paragraph after the share charts · methodology note \"LLM Market Share\"",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE12",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE12",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "Anthropic said in February 2026 that Claude Code's run-rate revenue had grown to over $2.5 billion.",
   "quote": "Today, Claude Code’s run-rate revenue has grown to over $2.5 billion",
   "removal_reason": "",
   "source": "Anthropic, Series G funding announcement, 12 February 2026",
   "source_url": "https://www.anthropic.com/news/anthropic-raises-30-billion-series-g-funding-380-billion-post-money-valuation",
   "locator": "paragraph beginning \"Claude Code represents a new era of agentic coding\"",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE13",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE13",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "DeepSeek publishes DeepSeek Harness (dsh) from its deepseek-ai GitHub organisation, and the README describes it as an open-source agent harness developed by DeepSeek AI.",
   "quote": "DeepSeek Harness (dsh) is an open-source agent harness developed by DeepSeek AI.",
   "removal_reason": "",
   "source": "DeepSeek AI, deepseek-ai/deepseek-harness README, GitHub",
   "source_url": "https://github.com/deepseek-ai/deepseek-harness",
   "locator": "README opening paragraph · DeepSeek's own site, deepseek.com/harness, links the repository",
   "date_read": "2026-10-03"
  },
  {
   "id": "HE14",
   "url": "https://durabilitycurve.com/claims/harness-engineering-same-model-different-product/#c-HE14",
   "status": "VERIFIED",
   "status_label": "Verified",
   "removed": false,
   "claim": "The DeepSeek Harness README labels the project a developer preview.",
   "quote": "DeepSeek Harness is in developer preview and iterating rapidly. THERE WILL BE COMPATIBILITY-BREAKING CHANGES.",
   "removal_reason": "",
   "source": "DeepSeek AI, deepseek-ai/deepseek-harness README, GitHub",
   "source_url": "https://github.com/deepseek-ai/deepseek-harness",
   "locator": "README section \"Developer preview\"",
   "date_read": "2026-10-03"
  }
 ]
}