{"slug": "proof-adjusted-autonomy-the-90-agent-is-a-61-6-agent", "title": "Proof-Adjusted Autonomy: The 90% Agent Is a 61.6% Agent", "summary": "A developer introduces Proof-Adjusted Autonomy (PAA), a metric that measures the share of autonomous work an organization can safely absorb after accounting for evidence, validation, and timeliness. Applying PAA to a demo agent with 90% raw autonomy yields only 61.6% effective autonomy, revealing a 28-point gap that becomes review queues and silent risk. The developer also defines Proof Debt as the accumulation of unverified AI-generated work, arguing that verification cost is the new bottleneck in production AI systems.", "body_md": "Every agent demo ends on the same slide: \"90% autonomous.\" Here is the number that slide is hiding: 61.6%.\n\nThe 90% is real. It measures how much work the agent completed without a human touching it. It just measures the wrong thing. Nobody runs a company on work that was *completed*. Companies run on work they can *accept* — without reconstructing it by hand to find out whether it's true.\n\n[Grant Thornton documented the problem this spring](https://www.grantthornton.com/insights/press-releases/2026/april/grant-thornton-survey-on-ai-proof-gap): organizations are deploying AI faster than they can demonstrate accountability for it. They call it the AI Proof Gap. [Jason Wei's Verifier's rule](https://www.jasonwei.net/blog/asymmetry-of-verification-and-verifiers-law) explains the deeper mechanism — the ease of training AI to solve a task is proportional to how verifiable the task is, which is why verifiable capabilities arrive first. But production systems face a third question neither of them answers: how much autonomous work can an organization safely absorb without checking it by hand?\n\n[Proof-Adjusted Autonomy](https://piszczek.pl/proof-adjusted-autonomy) measures that boundary.\n\nRaw autonomy counts tasks finished without human intervention. That's the demo metric. In production, every one of those tasks still has to pass four gates before anyone can act on it:\n\nProof-Adjusted Autonomy is the probability of passing all four:\n\n**PAA = P(A) × P(C|A) × P(R|A,C) × P(T|A,C,R) = P(A ∩ C ∩ R ∩ T)**\n\nEach factor is conditional on the previous gates, so the chain multiplies correctly — no independence assumption, no double counting. It's the actual share of your completed work that is autonomous, evidenced, validated and on time. You can estimate every factor from production logs.\n\nNow run the demo agent through it. Raw autonomy 0.90. Evidence coverage 0.80. Validation pass rate 0.95. On-time delivery 0.90. **0.90 × 0.80 × 0.95 × 0.90 = 61.6%.**\n\nMarketing reports the first factor. Operations lives with the product of all four. The 28 points between them didn't disappear — they became review queues, silent risk, and work a human quietly did twice.\n\nOne honesty clause, because the metric deserves it: P(R|A,C) must be estimated on a random or complete sample of evidence packages. If you only validate the work that's easy to validate, your PAA is a ceiling, not a measurement.\n\nAn agent can produce, in one hour: forty code changes, a two-hundred-page analysis, a thousand configuration decisions. Your organization still has to establish that the inputs were right, the goal was understood, the permissions were respected, the result works, and nothing else broke — and someone still has to sign.\n\nAI does not remove the cost of work. It [moves the cost from producing the work to proving the work is correct](https://piszczek.pl/blog/verification-cost-is-the-new-bottleneck).\n\nThis is where Verifier's rule cuts both ways. Wei is describing the learning frontier: what's easy to verify is easy to train, so capability floods into verifiable domains first. PAA describes the deployment frontier: whatever capability arrives, your organization can only operationalize the slice it can independently prove. The first frontier is set by the labs. The second one is set by you.\n\nAnd the second frontier compounds brutally. A fifty-step agent workflow at 99% per-step reliability completes cleanly 60.5% of the time. At 95%, it's 7.7%. Long-horizon agents don't primarily need smarter models. They need proof and correction at step boundaries — because reliability multiplies, it doesn't average.\n\nSo where do the missing 28 points go? They accumulate. Every piece of AI-generated work whose verification cost, uncertainty or liability hasn't been resolved yet is **Proof Debt**:\n\n**ProofDebt(t+1) = max(0, ProofDebt(t) + GeneratedWork − ProvenWork − RejectedWork)**\n\nIt's not just a review backlog. It's unproven assumptions, missing artifacts, decisions nobody can replay, and the future cost of reconstructing how something happened — payable on the day an incident, an audit or a customer claim asks the question.\n\nThis is the part your CFO should read twice. AI can raise reported productivity while silently accumulating Proof Debt. The P&L books the speed today. The incident books the liability later. A team that \"ships 3× faster\" with agents and no proof infrastructure hasn't tripled output. It has levered it.\n\nUnverified AI output is not an asset. It is deferred liability.\n\nAnd the debt has a hard ceiling behind it. If agents generate a hundred changes a day and your systems can independently prove thirty, your safe throughput is thirty — min(generation, verification), the oldest law in queueing. The other seventy aren't productivity. They're debt, accruing interest. Sustainable autonomy cannot exceed proof capacity.\n\nAt Archdesk we rebuilt our agentic engineering pipeline around this constraint. The agent's job doesn't end when it produces a result. It ends when the result survives independently defined acceptance.\n\nSo the agent never reports \"Fixed.\" It delivers an evidence bundle: the bug reproduced under a pinned configuration before the change; the diff and the operations log; tests passing; the same reproduction procedure demonstrating the corrected behaviour after; the neighbouring features checked for regression; the remaining uncertainty, stated; and a decision request for a human.\n\nThe before/after under an identical procedure is the part most teams skip — and it's the part that matters. A screenshot of a working page after the fix proves nothing; it would look identical if the fix were cosmetic. Evidence has to distinguish success from the appearance of success, or it's theater.\n\nOne design rule made most of the difference: *the agent never validates its own work*. A model grading itself shares its own blind spots, assumptions and error distribution. That's not independent verification — it's correlated confidence. Validation runs on different mechanisms: deterministic tests, replay, a different model family, a human wherever the action is irreversible.\n\nThe human role changes shape entirely. Reviewers stop reconstructing work and start adjudicating evidence. That's the whole economic point: review minutes per accepted task fall while PAA rises. We're instrumenting the pipeline now, and the numbers — raw autonomy versus PAA versus escaped defects, across model families — will be a separate publication. The framework is falsifiable, and it should be tested in public.\n\nIf PAA is the right lens, the next twenty-four months look like this. [Cost per verified task displaces cost per token](https://piszczek.pl/joule-wars) as the number that matters. QA stops being a phase and becomes the control plane of agent systems. Agent output stops being an answer and becomes an evidence bundle. The winning system won't be the one with the strongest model — it will be [the one that's cheapest to independently check](https://piszczek.pl/blog/who-owns-your-harness). Companies start reporting Proof Debt the way they report technical debt. Insurers and regulators start demanding replayability. And autonomy becomes a privilege agents earn with evidence history, not a toggle in a config file.\n\nWatch which of these happens first. That's the falsification schedule.\n\nIt is not the AI Proof Gap. Grant Thornton documented that the gap exists at enterprise scale — investment outrunning demonstrable accountability. PAA is the instrument: a number you compute from your own logs to measure the gap and watch it close.\n\nIt is not Verifier's rule. Wei's rule predicts which tasks AI will master fastest. PAA measures how much of that mastery your organization can let act. Learning frontier; deployment frontier.\n\nIt is not runtime verification research. Guardrails, evidence-bound execution and formal checking are mechanisms. PAA is the operational metric that tells you whether your mechanisms are actually buying you autonomy.\n\n## Key takeaways\n\n```\n- Raw autonomy measures work done without a human. PAA measures work done without a human that you can independently prove — and only the second number is deployable.\n- PAA is a chain of four conditional gates: autonomous × evidenced × validated × on time. 90% raw autonomy routinely collapses to ~60% PAA.\n- The gap between generated and proven work accumulates as Proof Debt — deferred liability that the P&L doesn't show until an incident prices it.\n- Safe throughput is min(generation rate, proof rate). Scaling agents without scaling verification scales debt, not output.\n- Self-verification is correlated confidence, not proof. Independence is what makes evidence evidence.\n- Sustainable autonomy cannot exceed proof capacity.\n```\n\nGeneration is no longer scarce. Proof is. The distance between them is where AI economics will be decided — and it's measurable. Measure it.\n\nThe canonical definition of [Proof-Adjusted Autonomy](https://piszczek.pl/proof-adjusted-autonomy) — and of Proof Debt — lives on its own page. Link it, argue with it, measure against it.\n\n*Originally published at piszczek.pl. The canonical definition of PAA and Proof Debt: piszczek.pl/proof-adjusted-autonomy.*", "url": "https://wpnews.pro/news/proof-adjusted-autonomy-the-90-agent-is-a-61-6-agent", "canonical_source": "https://dev.to/pich/proof-adjusted-autonomy-the-90-agent-is-a-616-agent-42jh", "published_at": "2026-07-14 23:21:27+00:00", "updated_at": "2026-07-14 23:57:40.649640+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-safety", "ai-agents", "ai-infrastructure", "mlops"], "entities": ["Grant Thornton", "Jason Wei"], "alternates": {"html": "https://wpnews.pro/news/proof-adjusted-autonomy-the-90-agent-is-a-61-6-agent", "markdown": "https://wpnews.pro/news/proof-adjusted-autonomy-the-90-agent-is-a-61-6-agent.md", "text": "https://wpnews.pro/news/proof-adjusted-autonomy-the-90-agent-is-a-61-6-agent.txt", "jsonld": "https://wpnews.pro/news/proof-adjusted-autonomy-the-90-agent-is-a-61-6-agent.jsonld"}}