The Arbiter
The Arbiter is a bullshit centrifuge for claims, drafts, papers, prompts, and AI-generated answers.
It does not exist to make writing smoother. Smooth writing is often the problem. The Arbiter tests whether a claim has met resistance: evidence, contradiction, consequence, limitation, method, source floor, and stake.
The basic question is simple:
Has this claim paid for what it says?
The Arbiter separates fluent language from earned language. It checks whether a text is supported by the right evidence floor, whether its definitions carry too much weight, whether its future claims are overpromised, whether its sources match its conclusions, and whether the user’s actual intent has been replaced by a familiar task.
Core review modes include Writing Collision, Paper Collision, Claim Collision, AI Output Trust Filter, Theory Collision, Evidence-Safe Revision, Prompt Collision, and Forward-Looking Statement Audit.
The point is not to punish ambitious writing. The point is to preserve ambition while forcing the claim to survive contact with reality.
Why this is different from asking ChatGPT
The Arbiter is not a claim that no AI was used. It is a claim about method.
A normal chatbot answer often optimizes for fluency, helpfulness, and completion. The Arbiter does something narrower and harsher. It forces a claim through a collision sequence: retrieval floor, load-bearing terms, evidence fit, contradiction, limitation, consequence, source floor, stake, and revision direction.
That structure matters. The goal is not to produce a smoother answer. Smoothness is often the problem. The goal is to separate what sounds convincing from what has actually paid for itself.
The Arbiter/Collider has been revised through repeated failures, patches, objections, source-floor repairs, overclaim checks, and hundreds of practical collisions. It is not a single prompt asking a model for an opinion. It is a pressure method built to make fluent language less safe, less frictionless, and more accountable to the claim it is making.
The current public version runs as a ChatGPT-based prototype and requires a ChatGPT account. A standalone Arbiter app is in development so users can run claims, papers, prompts, articles, market narratives, and AI answers through the method directly.
AI can assist the collision. It cannot replace the collision. The output still has to return to sources, limits, consequences, and human editorial judgment.
What can be collided?
The Collider is useful wherever fluent language can hide weak load-bearing structure.
Papers
Test abstracts, methods, claims, evidence floors, limitations, and conclusion strength.
Public arguments
Pressure-test slogans, headlines, theses, policy statements, and compact assertions.
AI answers
Check false certainty, missing source/date/jurisdiction, hallucination risk, and unsafe reliance.
Essays and theories
Find overreach, hidden assumptions, unsupported leaps, and stronger revision directions.
Market narratives
Not stock tips. Claim-pressure on forward-looking assumptions, risk, execution, margins, and sentiment.
Creative work
Test whether a script, pitch, concept, prompt, or visual idea has paid for what it claims.
Experimental theory lab
The Arbiter is the practical tool. The Seeing Loop Visualizer is an experimental research prototype for exploring the theory underneath it: occurrence, stake, descriptor, collision, verdict, action, consequence, and revision.
The visualizer is not a proof system and not required to use The Arbiter. It is a lab artifact for people who want to inspect the philosophical machinery.