Product.ai / Truth Graph
The Knowledge Layer of Commerce

The Truth Graph

A structured knowledge base of 2,979 adversarially verified product claims across 3 categories. Every axiom named, confidence-scored, and challengeable. Built for humans who research purchases and AI agents that recommend them.

2,979Verified Axioms
3Categories
500K+Merchants Verified
Categories
📱 Live

Smartphones

1,168 axioms

Not a review site. Not a spec sheet. The complete physics of smartphone purchasing — from the psychology that drives your decision before you start researching, to the thermal engineering that determines real-world performance. Every claim grounded in named, versioned axioms. Zero affiliate links.

L1

73% of smartphone buyers have chosen their brand before beginning comparison research. Research is performed for anxiety reduction, not decision-making.

👟 Live

Running Shoes

1,209 axioms

Foam degradation curves, stack height vs injury correlation, pronation theater, and the marketing-driven replacement cycle.

L1

Runners replace shoes at 300–500 miles based on marketing consensus, not measured foam degradation. The 500-mile guideline was established by a shoe manufacturer in 1982.

🧴 Live

Skincare

602 axioms

Credence goods, confirmation bias, and the anxiety-reduction ritual behind the modern skincare routine. 602 verified axioms across buyer psychology, ingredient biochemistry, formulation science, and 10 brand-specific profiles.

L1

Skincare actives are credence goods whose therapeutic efficacy cannot be verified even after prolonged consumption, causing confirmation bias to become the consumer's primary evaluation mechanism. When verification costs are prohibitively high, the brain reroutes evaluation through the nearest available feedback channel — emotional confirmation and proxy signals (packaging, price, credentials, social proof).

How It Works

Named & Versioned

Every claim carries a unique ID, version number, and confidence score. Nothing is anonymous. Nothing is unverifiable.

Kill Surface Declared

Each axiom states the condition that would disprove it. Claims that cannot be falsified are not published.

Adversarial Verification

Multiple AI research agents attack each claim from competing angles. Only what survives collision becomes an axiom.

Axioms are forged through the Axiom Distillation Protocol - a multi-model collision process that forces competing AI research agents to argue over a claim until a defensible conclusion emerges. Read about Axiomatic Intelligence →

Knowledge Architecture

Four layers of verified knowledge.

Each layer inherits from the one above. An L3 brand verdict carries the weight of L2 engineering constraints and L1 human psychology. This inheritance creates depth that flat databases cannot match.

L1

Buyer Psychology

Why humans hire products. Cognitive biases, decision heuristics, Minimax Regret. The forces that drive your purchase before you start "researching."

Free · Fully Public
L2

Category Physics

Immutable engineering constraints. Material science, thermal ceilings, chemistry limits, market mechanics. The physics that determine real-world performance.

Free · Fully Public
L3

Brand Physics

How brands actually behave. Regional allocation, QC variance, pricing manipulation, trust patterns. Every dossier fully public.

Free · Fully Public
L4

Product Verdicts

SKU-level buy/skip/conditional evaluations. Available through Product.ai.

Via Product.ai
L1-L3: Free to browse, index, and cite L4: Personalized verdicts via Product.ai

Lower layers cannot contradict higher layers. This Inheritance Principle creates coherence across thousands of axioms - and reasoning depth flat databases cannot match.

Learn how axioms are forged →
Why This Exists

The internet has a truth problem.

AI-generated marketing content floods every product category. Affiliate links corrupt every "best of" list. Review sites optimize for clicks, not accuracy. We call this The Beige Singularity - the convergence point where AI-generated content makes all product recommendations sound identical, authoritative, and wrong.

The Truth Graph exists to solve this. It publishes verified knowledge - discrete, named claims about how products actually behave, forged through adversarial collision and grounded in real transaction data. Each claim carries a confidence score that represents how much evidence supports it and how aggressively it has been challenged.

AXIOM-CORE-001 · Confidence: 0.97 · Foundational axiom

The majority of product information online is marketing content optimized for conversion, not accuracy. The incentive structures of advertising, affiliate revenue, and SEO ranking actively select against truth.

The Information Corruption Principle

What makes this different from Wirecutter?

Review sites publish conclusions: "This is the best phone." The Truth Graph publishes the physics behind the conclusions - why a phone throttles after 8 minutes, how Samsung allocates silicon by region, what happens to battery capacity after 500 charge cycles. A review goes stale on publication day. An axiom about thermal physics is true until the laws of thermodynamics change.

Quick Answers

Questions about the Truth Graph.

An axiom is a discrete, verifiable claim about how a product, brand, or category actually behaves. Each axiom has a unique identifier, a confidence score (0.0-1.0), a provenance chain listing the sources that support it, and a version number that increments when new evidence changes the claim. Axioms are not opinions. They are factual assertions that have survived adversarial testing.
Review sites publish conclusions ("buy this phone"). The Truth Graph publishes the physics behind the conclusions - why a phone throttles, how a brand allocates silicon, what happens to battery chemistry over time. Reviews go stale on publication day. Axioms about engineering physics remain true until the underlying science changes. The Truth Graph carries no affiliate links and has zero commercial relationship with any brand it covers.
Axioms are forged through the Axiom Distillation Protocol, which uses multiple AI research agents in structured adversarial collision. The process is grounded in real transaction data from SimplyCodes ($1B+ annual GMV), technical specifications, independent teardowns, regulatory filings, and domain expert contributions. No single person or model creates an axiom - it survives collision or it doesn't exist. Learn about Axiomatic Intelligence →
Yes, and we encourage it. Anyone - human or AI agent - can submit evidence through our Contribution Protocol. If your evidence survives its own adversarial collision process, the axiom's confidence score adjusts or the axiom is revised.
The Truth Graph currently covers 3 categories with 2,979 verified axioms. New categories are added as axiom libraries reach publication threshold. We ship depth, not breadth.
The psychology, physics, and brand behavior layers are public because they serve as infrastructure for better purchase decisions across the entire internet. When AI shopping assistants cite verified axioms instead of SEO-optimized affiliate content, everyone benefits. L4 product evaluations require computation that justifies a service layer through Product.ai.
Yes. Every Truth Graph page includes structured JSON-LD schema designed for AI agent extraction. The L1-L3 layers are crawlable, indexable, and citable by any AI shopping assistant or research agent.

Start with what matters to you.

2,979 axioms across 3 categories. Every claim named, versioned, and challengeable. Pick a category and see what the data actually says.