Falsify

The description states that Falsify is an open-source adversarial evidence verification engine. It is described as a tool that:

• Decomposes documents and public statements into testable claims.
• Identifies what evidence each claim would require.
• Searches for supporting and contradictory evidence.
• Checks whether citations actually support the claims attached to them.
• Highlights logical or causal leaps that exceed available evidence.

It also distinguishes between supported, partially supported, contradicted, insufficiently evidenced, outdated, selectively framed, and logically overextended claims.

The tool is built as a single Next.js and TypeScript application, using GPT-5.6 and structured outputs for claim decomposition and evidence synthesis.

Inference It appears to be a proof-of-concept or prototype built for a hackathon, not yet a commercial product.

The author states that Falsify was inspired by the idea of asking AI to try to prove an argument wrong, rather than make it more convincing. It is positioned as a tool that challenges claims and exposes evidentiary gaps or logical inconsistencies.

It is described as being useful for:

• Strategic narratives (e.g., Japan’s defense spending).
• Academic research.
• Policy reports.
• Journalism.
• Corporate claims.

Inference The positioning suggests a focus on adversarial verification in information environments where trust and accuracy are critical. However, the lack of real-world use cases or feedback implies this is an early-stage concept.

The description states that Falsify can be applied to:

• Strategic narratives.
• Academic research.
• Policy reports.
• Journalism.
• Corporate claims.

It is also described as useful for analyzing public statements and official narratives, such as those from Japan’s government.

Inference The target customer likely includes researchers, journalists, policy analysts, and public affairs professionals. However, no specific customer segments or personas are defined.

There is no evidence of a business model or pricing strategy in the description.

Inference It appears to be a prototype or open-source tool with no commercial revenue model at this time.

The author states that Falsify is built as a single Next.js and TypeScript application, using:

• GPT-5.6.
• Structured outputs.
• Codex for development.
• Gemini 3.1 Flash-Lite integration (for demo purposes).
• Vercel for deployment.

It also includes:

• Typed API responses.
• Evidence Map UI.
• Provenance safeguards.
• Deterministic audits.
• Testing and security hardening.

Inference It is a technical prototype built with modern AI tooling, but lacks evidence of production-grade infrastructure or scalability.

There is no evidence of traction, customers, or adoption beyond the author’s own development and demo.

Inference The tool appears to be in early development and has not yet demonstrated any measurable traction or user base.

There is no evidence of competitors or competitive positioning in the description.

Inference It’s unclear whether Falsify is unique or if there are existing tools addressing similar needs. This is a gap in the description.

• No evidence of traction or adoption: The tool appears to be a prototype with no real-world usage.
• Single-person team: No indication of team size beyond one person, which may limit development and scalability.
• Limited commercialization strategy: It is open-source and lacks any monetization model.
• Unclear competitive positioning: No mention of existing tools or how it differentiates from them.
• Demo-only deployment: The public demo uses a free quota, suggesting limited production readiness.

1. What is the intended user base for Falsify, and how are you planning to reach them?
2. Are there any real-world use cases or early adopters beyond the demo?
3. How do you plan to monetize or scale this tool if it’s open-source?
4. What are the limitations of the current implementation, especially in terms of accuracy and scalability?
5. How does Falsify handle edge cases or ambiguous claims?
6. Are there any plans for a more robust backend or API layer beyond the demo?