{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-20-003",
    "source_story_id": "tmp-story-mags-formal-agent-safety",
    "edition_id": "mp-2026-09-20-morning-0073",
    "edition_url": "https://themachinepress.com/edition/2026-09-20",
    "position": 3,
    "story_type": "dispatch",
    "section": "developer-tools",
    "editorial_classification": "editorial",
    "headline": "The Coding Agent Had to Bring a Proof",
    "slug": "the-coding-agent-had-to-bring-a-proof",
    "dek": "MAGS returned verified programs on all 220 tested kernels, scripts and robot-arm tasks under frozen specifications.",
    "summary": "MAGS returned verified programs on all 220 tested kernels, scripts and robot-arm tasks under frozen specifications.",
    "body_text": "MAGS turns human-audited APIs and safety requirements into frozen specifications, translates generated programs through Dafny, repairs verifier failures and compiles the verified result. Across 100 CUDA kernels, 100 terminal scripts and 20 robotic-arm tasks, every returned program carried a non-trivial machine-checked safety guarantee against those specifications. Independent evaluations still found failures when the automatically formalized semantics did not fully capture the intended behavior, so the guarantee is only as complete as the specification.",
    "why_it_matters": "MAGS returned verified programs on all 220 tested kernels, scripts and robot-arm tasks under frozen specifications.",
    "limitations": [
      "Independent evaluations still found failures when the automatically formalized semantics did not fully capture the intended behavior, so the guarantee is only as complete as the specification."
    ],
    "importance": 9,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-20-003/the-coding-agent-had-to-bring-a-proof",
    "json_url": "https://themachinepress.com/story/mp-2026-09-20-003.json",
    "first_published_at": "2026-09-20T09:00:00.000-04:00",
    "modified_at": "2026-09-20T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-20-003-001",
        "text": "MAGS returned verified programs on all 220 tested kernels, scripts and robot-arm tasks under frozen specifications.",
        "source_ids": [
          "source-2026-09-20-003"
        ],
        "qualification": "Independent evaluations still found failures when the automatically formalized semantics did not fully capture the intended behavior, so the guarantee is only as complete as the specification."
      }
    ],
    "source_ids": [
      "source-2026-09-20-003"
    ],
    "tags": [
      "coding agents",
      "formal verification",
      "Dafny"
    ],
    "image_url": "https://themachinepress.com/issues/2026-09-20/mags-code-file.webp",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-20-003",
      "title": "arXiv preprint 2609.19391",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.19391",
      "canonical_url": "https://arxiv.org/abs/2609.19391",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-16T16:15:36.000-04:00",
      "accessed_at": "2026-09-20T08:23:59.913-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-20-003-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The Coding Agent Had to Bring a Proof",
    "publisher": "The Machine Press",
    "published_at": "2026-09-20T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-20-003/the-coding-agent-had-to-bring-a-proof"
  }
}