{"slug": "i-open-sourced-a-macro-execution-layer-to-reduce-coding-agent-turns-60-task", "title": "I open-sourced a macro execution layer to reduce coding-agent turns (60-task benchmark)", "summary": "A developer open-sourced Tura, a macro execution layer for coding agents that reduces the number of model turns by executing routine workflow steps together. On a 60-task DeepSWE benchmark, the macro approach with backward reasoning achieved an 80% pass rate with 2,017 rounds, compared to Codex CLI Medium's 63.3% pass rate with 3,140 rounds. The project aims to lower orchestration turns without hiding failures, though the developer notes that fewer turns do not guarantee lower end-to-end cost.", "body_md": "**Disclosure:** I maintain Tura.\n\nA coding agent often spends a separate model turn on each part of a routine workflow: inspect the environment, edit package files, patch the implementation, update tests, build, run lint/tests, then inspect Playwright media.\n\nTura experiments with a simpler execution model: describe that sequence as a macro and let the runtime execute the steps together, while still returning structured results to the agent. The goal is to reduce orchestration turns without hiding failures.\n\nOn our current 60-task DeepSWE set:\n\n| Configuration | Passes | Pass rate | Observed tokens | Rounds |\n|---|---|---|---|---|\n| Macro + backward reasoning | 48/60 | 80.0% | 229,695,477 | 2,017 |\n| Macro Direct | 39/60 | 65.0% | 75,108,167 | 969 |\n| Codex CLI Medium | 38/60 | 63.3% | 333,538,349 | 3,140 |\n| Codex CLI High | 36/60 | 60.0% | 455,742,296 | 6,074 |\n\nThe implementation, task data, and benchmark methodology are public:\n\nThe important caveat is that fewer turns do not automatically mean lower end-to-end cost. Retries, failures, cache behavior, and completed-task rate all belong in the denominator. I would especially appreciate feedback on the benchmark design and cases where macro execution makes debugging worse rather than better.", "url": "https://wpnews.pro/news/i-open-sourced-a-macro-execution-layer-to-reduce-coding-agent-turns-60-task", "canonical_source": "https://dev.to/yohjisakamoto/i-open-sourced-a-macro-execution-layer-to-reduce-coding-agent-turns-60-task-benchmark-3o41", "published_at": "2026-07-18 22:44:18+00:00", "updated_at": "2026-07-18 23:29:26.242884+00:00", "lang": "en", "topics": ["developer-tools", "artificial-intelligence", "ai-agents", "machine-learning"], "entities": ["Tura", "DeepSWE", "Codex CLI"], "alternates": {"html": "https://wpnews.pro/news/i-open-sourced-a-macro-execution-layer-to-reduce-coding-agent-turns-60-task", "markdown": "https://wpnews.pro/news/i-open-sourced-a-macro-execution-layer-to-reduce-coding-agent-turns-60-task.md", "text": "https://wpnews.pro/news/i-open-sourced-a-macro-execution-layer-to-reduce-coding-agent-turns-60-task.txt", "jsonld": "https://wpnews.pro/news/i-open-sourced-a-macro-execution-layer-to-reduce-coding-agent-turns-60-task.jsonld"}}