Launching today
Lise Science
Ask a question, receive science
11 followers
Ask a question, receive science
11 followers
Lise is an autonomous research engine. It reads the literature, forms hypotheses, runs the numbers, and returns a report where every claim is traceable to its evidence. Lise Meitner, 1878–1968, she explained nuclear fission. We named Lise after her.
Love the nod to Lise Meitner and the emphasis on traceable claims. One thing that would help a lot: let users upload their own PDFs or supplementary datasets and have Lise cross-reference those against its literature findings, so the reports blend published evidence with whatever internal notes or unpublished data you bring to the table.
@aryaarkayaulwb you can do that! the base of this is that you have 5 running modes: quick, light and deep are for reference reading or dataset reading, like for instance gettin the state of the art on a question. Then Scientist and Program mode are fully autonomous research agents that take your data, your own findings and everyhting you add and perform your research goals. Program mode can take up to 24 hours of computing, going so deep that you can really adavance your research by months in hours.
Being able to save searches and get a notification when new papers come out that match my hypothesis would be huge. Right now it feels like each run is a one-off, but a lot of research questions evolve over time as literature shifts.
@necdetc2xx The base of this is that you have 5 running modes: quick, light and deep are for reference reading or dataset reading, like for instance gettin the state of the art on a question.
Then Scientist and Program mode can be fully autonomous of human in the loop research agents that take your data, your own findings and everyhting you add and perform your research goals. Program mode can take up to 24 hours of computing, going so deep that you can really adavance your research by months in hours.
Once they finish, you can get another program to run from there, and you can be the human in the loop giving directions and correcting things on the fly. The goal is that for example if you are solving a problem related to proteins and you wanna delegate all the computation, you can do it here and you can get the results at the end. I created this to speed up research on my own battery startup, and it was so useful that now Lise is a product itself. Please use it and ask me what things would you like me to add.
the traceability angle is genuinely cool, big fan of that. one thing i'd love is a way to export the underlying evidence graph as a structured file so i can drop it into zotero or share the reasoning chain with collaborators without screenshots. would make it way easier to cite or audit.
@mirayj8ur will add that! Right now you can export everything, but will add a graph to make it easier with a direct zotero link.
Gave it a spin on a tricky metabolism question and the traceable evidence trail was genuinely useful, saved me an hour of digging through PubMed. Love that it's named after Lise Meitner too.
@naimekini8upz thank you!