Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/XxYouDeaDPunKxX/PA-PVP-mini/llms.txt

Use this file to discover all available pages before exploring further.

Two-AI mode turns PA-PVP Mini into a ping-pong between two completely separate AI chat sessions. AI-A reviews your artifact and returns Round 1 findings. You then open AI-B, activate the same protocol, and give it both the original artifact and AI-A’s full output. AI-B independently reviews the artifact first, then audits AI-A’s findings — confirming what held up, challenging what didn’t, and surfacing anything that was missed. You can continue alternating between the two sessions for as many rounds as useful.

​
What each AI does

Every AI in the chain does two things in strict order:
  1. Independent review — reviews the ORIGINAL_ARTIFACT without looking at the prior output first
  2. Prior-round audit — cross-checks the previous AI’s findings against its own independent review
This separation is what makes the ping-pong valuable. Because each AI completes its own review before examining the prior one, agreements are genuine confirmations rather than echo-chamber repetition.

​
Setup flow

1

Activate PA-PVP Mini in AI-A

Open a new chat session with your first AI. Paste the full contents of pa_pvp mini v2.3.txt to activate the protocol.
2

Submit your artifact and receive Round 1

Send your material labeled as ORIGINAL_ARTIFACT:
ORIGINAL_ARTIFACT:
[your material]
AI-A returns Round 1. All findings are labeled INDEPENDENT. Copy the entire output — you will need it in the next step.
3

Open AI-B and activate PA-PVP Mini in a new chat

Open a completely separate chat session with your second AI (or a new session with the same AI provider). Paste pa_pvp mini v2.3.txt again to activate a fresh protocol instance with no shared memory.
4

Submit the artifact plus AI-A's Round 1 output to AI-B

In the AI-B chat, send both blocks together:
ORIGINAL_ARTIFACT:
[your material]

PREVIOUS_ROUND_OUTPUT:
[AI-A Round 1 output]
AI-B first produces its own independent review of the artifact, then audits AI-A’s findings, labeling each as CONFIRMED, CHALLENGED, or NEW.
5

Review the Round 2 output from AI-B

AI-B’s output includes a MERGE STATUS block that summarises the state of all findings across both rounds. Use it as a triage guide before deciding what to act on (see Applying Fixes for details).
6

Continue by copying AI-B's output back to AI-A

To run Round 3, go back to the AI-A chat (or open a fresh session with AI-A), and send the original artifact together with AI-B’s complete Round 2 output as PREVIOUS_ROUND_OUTPUT. AI-A now audits AI-B’s findings. Continue alternating for as many rounds as needed.

​
No shared memory required

Each round is fully self-contained. The full text of PREVIOUS_ROUND_OUTPUT carries all the context the next AI needs — finding IDs, evidence, fixes, severity labels, and provenance. Neither AI needs access to the other’s chat history. You can even run the two sessions days apart.

​
MERGE STATUS fields

Round 2+ outputs include a MERGE STATUS block at the end of the findings. The fields are:

confirmed_fixes_ready ✅

Fixes from the prior round that survived the current independent review. These have been validated by two separate passes.

challenged_items 🥊

Prior findings or fixes that the current AI disputes — because they are invalid, overstated, duplicated, weak, or not anchored to the artifact.

new_fixes 🆕

Fixes for problems found in the current round that were not present in any prior round. These are fresh and have not yet been cross-checked.

conflicts ⚠️

Direct contradictions between the current independent review and the prior audit position. Conflicts are declared explicitly rather than silently resolved — they require human judgment.
Use two different AI providers — for example ChatGPT and Claude — rather than two sessions with the same provider. Different training, different tendencies, and different knowledge weighting produce genuinely divergent findings. Agreements between them carry more weight; contradictions surface assumptions worth examining.
Always include the full ORIGINAL_ARTIFACT in every pass — not just the previous round’s output. PREVIOUS_ROUND_OUTPUT is audit material, not source authority. An AI given only the prior review has no artifact to anchor its findings to and cannot produce valid evidence.
For details on how PA-PVP Mini auto-detects the current round number from the blocks you supply, see Round Detection.

Build docs developers (and LLMs) love

Get started for freeTalk to us