Agents In UseFutureHouse ran WikiCrow to draft scientific summaries of human genes using full-text papers. Its PaperQA agent searched, gathered evidence and returned to retrieval when terminology or evidence required it.
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
Using the accessible papers for [gene], draft a short research summary. Cite each substantive claim, separate direct evidence from inference and flag contradictions. Return to retrieval when evidence is missing. Produce a draft for expert review, not a publication-ready clinical reference.
FutureHouse reports producing draft gene articles and finding incorrect statements in its evaluation; drafts remain fallible.
Reported by FutureHouse / Sam RodriquesResearchers audited sampled statements and warned readers to check claims against their citations.
A person approved each step before the agent continued.
PaperQA adaptively chose search and evidence tools, including new searches from intermediate findings.
Primary research report of a completed corpus run. These were drafts, not a claim of accepted Wikipedia publication.
Credit: FutureHouse / Sam Rodriques
Source published 8 Dec 2023
Source checked 14 Oct 2026
WikiCrow · PaperQA · SEARCH · GATHER_EVIDENCE · ANSWER_QUESTION · Semantic Scholar API
Research · Sources
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.
til.simonwillison.netWillison ran o4-mini-deep-research to find surviving orchestrions and return structured location records. The model used web search and code execution; he explicitly considered the resulting list incomplete and unverified.
Added
OpenAI deep research
futurehouse.orgFutureHouse used Robin to connect literature research, experiment planning and analysis for dry age-related macular degeneration. Human researchers ran the lab work; the system proposed follow-up candidates from the results.
Added
Robin