About Adversary Loops
Most AI writing tools help you write more. Adversary Loops helps you write better.
Inspired by Andrej Karpathy's autoresearch method, Adversary Loops puts every document through an adversarial gauntlet. An AI critic attacks your writing — finding weak claims, vague numbers, logical gaps, and unsupported assertions. Then a drafter fixes what the critic found. Then a regression checker makes sure nothing got lost in the rewrite.
This cycle repeats — round after round — until there's nothing left to attack.
How It Works
- 1
Paste your text
Submit any high-stakes document — an investor update, cold email, blog post, research paper, or product launch announcement.
- 2
Critic attacks
An AI critic reviews your text and identifies every weakness: vague claims, missing evidence, logical inconsistencies, and structural issues.
- 3
Drafter fixes
A separate AI model rewrites your text to address every issue the critic found, while preserving your voice and intent.
- 4
Regression check
A verification pass ensures nothing important was lost or distorted during the rewrite — factual assertions, key data points, and core arguments are all preserved.
- 5
Repeat until bulletproof
The loop continues until the critic can no longer find substantive issues, or improvements converge. Your text emerges sharper, more precise, and harder to argue with.
Built for High-Stakes Writing
Adversary Loops is designed for documents where precision matters — investor updates that need to be airtight, proposals that need to persuade, emails that need to convert, and papers that need to withstand scrutiny.
Multiple AI Models
Choose from leading AI models including GPT-4o Mini, GPT-4o, Claude Sonnet 4, and Gemini 2.0 Flash. Free-tier users get access to efficient models, while Pro users unlock the most capable models available.