Traditional knowledge systems treat information as static. You write something, it gets published, and there it sits until someone decides to update it. AI Immortal takes a different approach: knowledge is dynamic, contested, and always provisional.

Claims: The Atomic Units

Everything in AI Immortal starts with claims. A claim is a single, verifiable assertion. Not a vague statement. Not an opinion. A specific claim about reality that can be supported or challenged with evidence.

Good claims are:

- Specific and falsifiable

- Verifiable against evidence

- Properly scoped in time and domain

- Connected to sources

Bad claims are:

- Vague or ambiguous

- Unfalsifiable

- Opinion masquerading as fact

- Disconnected from evidence

The AI Immortal system extracts claims from text automatically but also lets humans refine them. Getting the claim boundaries right matters because everything else builds on them.

Evidence: The Foundation

Claims without evidence are just assertions. Evidence gives claims strength. But not all evidence is equal.

We track:

**Source Trust**: Has this source been reliable in the past? Do they have domain expertise? Have previous claims citing this source held up or failed?

**Evidence Type**: Primary sources trump secondary. Direct data beats interpretation. Recent evidence generally outweighs old.

**Independence**: Multiple independent sources strengthen claims. Five news articles all citing the same study don't count as independent evidence.

**Recency**: Evidence has a freshness date. A 2020 study about technology trends is less relevant in 2026 than a 2025 update.

Each evidence link includes metadata: what type of evidence it is, when it was published, who authored it, and what trust score the system assigns. Users can see this metadata and challenge it.

The Dispute System

This is where it gets interesting. Any claim can be disputed. When you dispute a claim, you're not just leaving a comment. You're triggering a formal evaluation process.

**Step 1: Dispute Creation**

You specify what's wrong: factual error, misrepresented evidence, outdated information, missing context, or flawed reasoning. You provide counter-evidence.

**Step 2: Response Window**

The original claimant gets a chance to respond. They can provide additional evidence, clarify the claim, or acknowledge the error.

**Step 3: Evaluation**

If both sides present arguments, an AI judge evaluates them using the standard rubric. The judge decision is transparent and challengeable.

**Step 4: Resolution**

Winning disputes lead to claim updates. The system tracks dispute history so you can see how claims evolved. Losing disputes don't disappear but remain visible as attempted challenges.

Why Disputes Matter

Disputes serve multiple purposes:

**Error Correction**: The most obvious benefit. Wrong claims get challenged and fixed.

**Trust Scoring**: Users and sources build reputation through dispute outcomes. Consistently accurate claims and evidence sources gain trust.

**System Learning**: Disputes that expose judge errors help improve the AI evaluation system.

**Explicit Uncertainty**: Active disputes signal that a claim is contested. Readers see the controversy, not just the current winning position.

Cascading Updates

Claims often depend on other claims. When a foundational claim gets successfully challenged, dependent claims may need revision. The system flags these cascading implications.

For example, if Claim A uses Claim B as evidence, and Claim B gets overturned in a dispute, Claim A gets flagged for review. This prevents outdated dependencies from hiding.

The Meta Game

Savvy users recognize that the dispute system isn't just about correcting individual claims. It's about building reputation, finding weak points in the knowledge graph, and improving the system itself.

Good disputers:

- Choose important, widely-cited claims to challenge

- Provide clear, well-sourced counter-evidence

- Engage with the strongest version of opposing arguments

- Learn from dispute outcomes

Poor disputers:

- Nitpick trivial details

- Rely on weak or biased sources

- Misrepresent opposing positions

- Ignore rubric feedback

Future Directions

We're exploring:

- Prediction markets tied to disputes (bet on outcomes)

- Dispute escalation paths for high-stakes claims

- Community voting on dispute importance

- Automated dispute suggestion when new contradictory evidence emerges

The goal is a living knowledge system that improves through challenge. Truth isn't static. It's the current best answer that has survived the most rigorous scrutiny.

Get Involved

See a claim that seems wrong? Dispute it. Have strong evidence for a controversial position? Submit it. The system gets stronger with participation.

Remember: disputes aren't personal attacks. They're how knowledge gets better. Good claims survive challenges. Weak claims get refined or replaced. Everyone learns.