Traceable support
Important claims stay linked to relevant source material. URL and title alone are not evidence; the excerpt must support the claim being made.
How it works
Novelty Engine is a research and elimination system. It organizes evidence, tests what that evidence can support, and leaves uncertainty visible instead of filling gaps with confident language.
The process
Name the market, customer, problem, company, or idea you want to investigate.
Claude or you collect bounded excerpts from relevant public sources and keep each source URL and title attached.
Novelty Engine checks competitors, recurring pain, workarounds, source quality, counterevidence, and missing facts.
Only supported conclusions survive. The report keeps the decisive risk, unknowns, sources, and a practical validation step visible.
Supplied-source research
The normal workflow uses sources supplied by Claude or by you. Each item includes a public URL, a title, and a bounded evidence-bearing excerpt. Novelty Engine analyzes that package without calling Tavily or Brave.
Optional hosted search remains available only when a deployment enables it and a user explicitly requests it. The report preserves which retrieval path was used, so supplied evidence is never described as Novelty Engine having searched the web.
Evidence requirements
Important claims stay linked to relevant source material. URL and title alone are not evidence; the excerpt must support the claim being made.
Repeated pages from the same publisher do not become independent proof. Source role, company identity, and publisher identity stay separate.
A weak idea is rejected when the pain is isolated, competitors already solve the job, the evidence contradicts the claim, or the decisive facts remain missing.
Missing facts remain unknown. Rejected conclusions and reasons for rejection stay in the record instead of disappearing from the final answer.
Public discussions
You or Claude may provide bounded excerpts or public thread content from Reddit and other lawful public research sources. Novelty Engine does not directly scrape, crawl, or mine Reddit, and a supplied discussion is treated as a limited sample—not complete forum coverage.