# Let a team of AI agents propose a new treatment idea

FutureHouse's Robin proposed ripasudil, a glaucoma drug, as a candidate for dry macular degeneration.

Organisation: FutureHouse
Status: Research (published in Nature)
Place: United States
Categories: Science & discovery, Health & care
Built with: FutureHouse agents
Try it with: Claude, ChatGPT / OpenAI models, Gemini, Grok
Time to try: 1 to 2 hours with a deep-research agent

## Why it matters

Dry age-related macular degeneration is a leading cause of blindness. Agents that read the literature, plan experiments and analyse results can shorten the path to candidates worth testing.

## How it works

Robin orchestrates FutureHouse's literature agents (Crow, Falcon) and data-analysis agent (Finch). It proposed boosting retinal pigment epithelium phagocytosis, suggested candidate molecules that humans tested in the lab, analysed the results and later proposed an RNA-seq experiment that pointed to ABCA1. The work is published in Nature. The results are in cell culture, not in patients.

## Who it's for

Researchers, biotech founders and research funders.

## The prompt

```
Act as a research team of three agents working on [disease]:
- Literature agent: summarise the main disease mechanisms and the evidence for each, with citations (author, year, journal, DOI) only for papers you have actually retrieved with search. If you cannot search, give no DOIs: mark each claim [CITATION NEEDED] and list search terms instead.
- Hypothesis agent: propose 5 therapeutic strategies, each linked to the evidence above.
- Critic agent: score each strategy from 1 to 5 for novelty, evidence and testability, and name the main reason it could fail.

Then rank the top 3 and, for each, propose one cell-based experiment: model, readout, controls and what result would falsify it.

Flag any claim you are not sure of. Do not present ideas as medical advice.
```

## Caveats

Candidate validated in cell culture only.

## Sources

All three sources are FutureHouse's own work: its site, its peer-reviewed Nature paper and its arXiv preprint. No independent evaluation was found.

1. FutureHouse: Crow, Falcon, Finch; ripasudil for dAMD; ten molecules tested. Demonstrating end-to-end scientific discovery with Robin (https://www.futurehouse.org/research/demonstrating-end-to-end-scientific-discovery-with-robin-a-multi-agent-system)
2. Nature: Peer-reviewed: ripasudil and KL001 in vitro; ABCA1 upregulation. A multi-agent system for automating scientific discovery (https://www.nature.com/articles/s41586-026-10652-y)
3. arXiv: Preprint abstract. Robin (arXiv preprint) (https://arxiv.org/abs/2505.13400)

Fact-checked: 16 Oct 2026

Page: https://escapevelocitysociety.com/picks/futurehouse-robin-drug-candidate

See Selected Projects: https://escapevelocitysociety.com/picks.md
