WorthDoing

WorthDoing.ai answers a different question from search engines: what should exist, be built, researched, tested, funded, or pursued that does not exist yet — or is not being pursued enough? It is not a thin LLM wrapper and not a fixed query → search → scrape → summarize pipeline: Claude repeatedly decides what it knows, what remains uncertain, what to search next, which sources to inspect, whether to reject hypotheses, and when the evidence suffices to conclude.

Every investigation runs through five phases. Scout maps a domain with varied Firecrawl searches and forms falsifiable hypotheses. Investigate runs targeted searches (verify / market / technical / competition) and saves evidence as near-verbatim quotes linked to hypotheses. Skeptic is mandatory falsification — every high-confidence hypothesis gets adversarial falsify searches, and hypotheses weaken, branch, or die; a rejected hypothesis is useful progress. Synthesize turns survivors into opportunities with a seven-dimension, evidence-cited Worth Score (demand, neglectedness, feasibility, why-now, impact, competition, risk), with evidence confidence reported separately, never blended in. Report streams a citation-grounded report per opportunity, where citations resolve mechanically to saved evidence.

The guarantees are hard, not aspirational: budgets are hard bounds (30 steps, 20 searches, 60 scrapes, 3 crawls, wall-time) with structured stop reasons; no conclusion ships without traceable evidence — every score dimension carries {score, confidence, reasoning, evidenceIds}; scraped content is untrusted data, never instructions; and investigations are resumable because state lives in Postgres, not in the model's context. Every UI event corresponds to a real backend AgentEvent streamed over SSE — the live timeline never fabricates progress.