ADVERSARIAL REVIEW · HUMAN ADJUDICATION · THE DOSSIER

Models diverge. Answers converge.The ruling is yours — on the record.

Ask one model to check its own work and you get a more articulate yes-man. Parry sends your question to rival frontier models that reason independently, then cross-examine each other. The disagreements that survive become your call — and the whole deliberation is kept as a dossier you could defend months from now.

NO WAITLIST · FIVE BOUTS FREE · BYOK — PAY PROVIDERS AT COST
WHY A RIVAL MODEL — NOT THE SAME ONE AGAIN

Ask a model to double-check its own work. You get a more articulate yes-man.

TRAINED TO AGREE

Approval is optimized for. Agreeable and correct are different targets.

SELF-REVIEW IS A MIRROR

Same weights, same blind spots, same incentives — review in name only.

THE FAILURE IS INVISIBLE

After “review,” the answer reads more confident — not more correct.

AND NO RECORD → When an AI-assisted call goes wrong, the three questions every post-mortem asks — what was it challenged on, who decided to accept it, what would have changed it — have no answer.

THE FIX IS AN OLD IDEA

Every serious field reviews consequential claims adversarially — on the record.

SciencePEER REVIEW
LawOPPOSING COUNSEL
MedicineTHE SECOND OPINION
SoftwarePULL-REQUEST REVIEW
AI outputASKED TO REVIEW ITSELF

Independent models from different vendors cross-examine each other. Surviving disagreements become explicit human decisions. The whole deliberation is memorialized as a dossier.

Parry removes the exemption — convergence you can defend.
EXHIBIT A · A REAL DOSSIER, UNEDITED
PALLETS/FLASK#5898 — A PUBLIC PR
ANTHROPIC CLAUDE OPUS 4.8claude-opus-4-8OPENAI GPT-5.5gpt-5.5FLOW 2 · 6 TURNS · 3M 09S · $0.91 · 2026-07-07
THE QUESTION
Should pallets/flask#5898 merge? The PR changes Flask's redirect() default status code from 302 to 303.
VERDICTREVISECONFIDENCE MEDIUM-HIGH
Merge after fixing the stale 308 “can't preserve form data” warning in docs/api.rst; the 302→303 default change is sound and intentional.
CONTENTIONS · 23 DUAL-CONFIRMED · 0 CONTESTED · 0 DISMISSED
§1LOWMissed test-coverage / CI-break risk for changed default
Claude: Their review: only issue found is the stale doc sentence; merge-with-fixes, high confidence — no mention of tests. GPT-5.5: No test file in diff or manifest; if the existing suite asserts the default redirect status is 302, this PR breaks CI. OPEN
§2LOWMissed explicit API/contract-shift callout (302→303)
Claude: Scope mentions callers now emit 303, but no finding records the contract shift or its non-browser-client implications. GPT-5.5: 303 differs from 302 in cacheability and forces method→GET for non-browser HTTP clients. Deserves an explicit contract-risk finding. OPEN
THE FULL RECORD: 2 CONTENTIONS · 6-TURN TRANSCRIPT · 123,841 TOKENS · COST RECEIPTSSIGN UP TO RUN YOURS — SHARE LINKS SHIP NEXT

This is the artifact, not a mock of one — A real review of a public pull request — pallets/flask#5898 — run through Parry's PR-review pipeline. Go read the PR yourself.Every Parry run produces a page like this. When the models agree, it says so; when they don't, the disagreement is the point.

WE EAT THIS DOGFOOD
Parry reviews its own commits before they ship.

The parser fix in this very build was cross-examined by rival models the day it shipped — verdict: merge as-is, both reviewers independently confirmed no defects, with one honest risk left on the record (“downstream consumer unverified — withheld”).

RECOMMEND · CONFIDENCE HIGH · $0.86 · 134,824 TOKENS · 1M 44S
HOW A BOUT WORKS

Not a second opinion. A cross-examination.

01

Propose

Each model commits to a position independently — specific, falsifiable, trade-offs named. No peeking at each other.

02

Cross-examine

They attack each other's work. Concessions are recorded; positions that hold under pressure are recorded too. Sycophancy is structurally penalized.

03

You rule

Disagreements that survive become explicit decision points. You rule on them — and your ruling is part of the record, with your name on it.

04

On the record

Verdict, contentions, transcript, receipts — memorialized as a dossier. Six months later you can show exactly why the call was made, and who made it.

THE HONEST CLAIM → Heterogeneous models notice different things — that's the research finding we're built on. We don't claim better answers. We claim you'll see the disagreement before you commit, instead of after.

FOR ENGINEERS · GITHUB ACTION · BYOK · IN DEVELOPMENT

Your PRs, cross-examined. One comment. Silence when they agree.

Two models review every PR independently, then challenge each other. You get a mini-dossier comment only when disagreement survives — with both positions and a checkbox for your ruling. No walls of nitpicks. Your keys, your spend — $0.91 a review, measured on real PRs (n=2 — small sample, run-to-run variance is real).

THE ACTION IS SHIPPING SOON — THE REVIEW PIPELINE BEHIND IT IS LIVE (EXHIBIT A RAN ON IT) · ALSO COMING: MCP SERVER FOR CLAUDE CODE / CURSOR
# .github/workflows/parry.yml — target config
- uses: parry/review-action@v1
  with:
    models: claude-sonnet-5, gpt-5.5
  env:
    ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_KEY }}
    OPENAI_API_KEY: ${{ secrets.OPENAI_KEY }}
SAMPLE COMMENT · TARGET FORMAT
⚔️ Parry adversarial review — REVISE · 2 disagreements survived. §1 — test coverage: no test asserts the new default; if the suite pins 302, this breaks CI. ☐ Uphold Claude ☐ Uphold GPT-5.5 ☐ Discuss · Full dossier →

10 domains. Same adversarial method.

EACH SHIPS WITH TEMPLATES TUNED TO HOW THAT FIELD ARGUES
SOFTWARE DEVELOPMENTBUSINESS STRATEGYRESEARCH ANALYSISPRODUCT DEVELOPMENTDEBATE & ARGUMENTATIONPRODUCT RESEARCHMEDICAL OPINIONSLEGAL REVIEWINVESTMENT ANALYSISWRITING CRITIQUE+ DOMAINS ARE CONFIG — THE NEXT ONE IS ONE REGISTRY ENTRY

Sensitive domains (medical, legal, investment) carry explicit disclaimers and honest labels — analysis for deliberation, not professional advice. That's a feature of the record, not fine print.

Pricing that reads like a receipt.

NO SEATS THEATER · NO "CONTACT SALES"
FREE
$0
five bouts on our keys
  • All domains, all three flows
  • Full dossiers, MD export
  • No card required
BYOK — THE HONEST TIER
At cost
your keys · pay providers directly
  • Unlimited bouts on Anthropic / OpenAI keys
  • GitHub Action + MCP server (in development)
  • Keys encrypted at rest; every token metered and itemized
PRO
$20/mo
the record-keeping layer · billing opens with the first cohort
  • Hosted dossier history & private sharing
  • PR ruling sync — checkboxes become dossier rulings
  • Team spaces, later — you'll hear it from us first

The inference is cheap and getting cheaper — we don't mark it up. What you pay for is the record: history, sharing, and rulings that survive.

The next hard call you make — put it on the record.
Sign up, ask your question, read the disagreement. Five bouts free.
Start a bout →
Parry. PARRYAITOOL.COMPRIVACY · TERMS · © 2026 M3 LABS LLC · PATENT PENDING