{"slug": "evidence-runtime-for-autonomous-agent-task-management", "title": "Evidence runtime for autonomous agent task management", "summary": "Grainulator, an evidence runtime for autonomous agent task management, enables AI models to incorporate independent checks and recorded evidence into their workflows, with a playground and local command-line tool that reproduces an evidence freshness bug fix and supports sending failed checks back to an adapter for retries. The tool, available via a session export and local checkout, does not establish a general accuracy or speed advantage based on local comparisons on Astra and Fable.", "body_md": "Research and verification for AI work\n\n# Keep your model.\n\n*Raise the standard.*\n\nBring evidence, independent checks, and useful corrections into the work your model already does. Keep the reasoning behind the result, so you can pick it up, question it, and move it forward.\n\n[Open the playground](/playground/)\n\nSee what a check catches\n\n## Good yesterday.\n\nStill good today?\n\nA check passes. The work moves on. Then the evidence changes—and the old approval no longer holds. Explore a real bug we fixed in Grainulator: catching that mismatch before a stale brief goes out.\n\nFollow the evidence\n\n### Inspecting the recorded check…\n\nThe trace is a small local file. There are no model requests or timed animations.\n\n—\n\n- Checked content\n- —\n- Current content\n- —\n\nWhere Grainulator fits\n\n## A stronger process.\n\nWith your existing tools.\n\nYour model brings the intelligence. Grainulator brings a record of the evidence, checks you define, and a way to turn failures into the next useful step.\n\n1. 01### Start with the workBring your question, task, and constraints. Keep the model and workspace you already use.\n2. 02### Make “done” testableChoose an independent test or verifier. See what passed, what failed, and what remains uncertain.\n3. 03### Give the next attempt a reasonFeed failed checks back to the model. Set limits on retries and keep the record of each attempt.\n\nFrom playground to real work\n\n## Find your setup.\n\nTake it with you.\n\nChoose your model and research settings in the playground. Export the session, then continue from your local checkout with the context and completed passes intact.\n\n```\nnode bin/grainulator.js research --session session.json --dir ./research-session\n```\n\nDownload session.json from the playground, then run this from your local checkout. Uses your provider key and continues the remaining passes.\n\n## What you can verify today\n\nThe trace reproduces an evidence freshness bug and verifies that the updated guard blocks stale output. The command runner can send failed checks back to an adapter for another attempt. Support for a particular model depends on that adapter or host.\n\nLocal comparisons on Astra and Fable measured answer correctness, formatting, elapsed time, and model usage. Some runs improved formatting; the results do not establish a general accuracy or speed advantage. The verifier rejected candidates that failed its checks.\n\nReproduce the trace with `node scripts/record-handoff.mjs`. The adapter contract is in `docs/ADAPTERS.md` in the dogfood checkout.", "url": "https://wpnews.pro/news/evidence-runtime-for-autonomous-agent-task-management", "canonical_source": "https://grainulator.app", "published_at": "2026-09-08 04:24:15+00:00", "updated_at": "2026-09-08 05:02:12.682496+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "ai-infrastructure"], "entities": ["Grainulator", "Astra", "Fable"], "alternates": {"html": "https://wpnews.pro/news/evidence-runtime-for-autonomous-agent-task-management", "markdown": "https://wpnews.pro/news/evidence-runtime-for-autonomous-agent-task-management.md", "text": "https://wpnews.pro/news/evidence-runtime-for-autonomous-agent-task-management.txt", "jsonld": "https://wpnews.pro/news/evidence-runtime-for-autonomous-agent-task-management.jsonld"}}