Unlocking liquid science markets through tokenized IP
- Tweets 6,744
- Following 2,020
- Followers 30,369
- Likes 9,474
Molecule retweeted
AI workflows that keep the scientist in the loop was Molecule's table at AI in Life Sciences, run by @peptai_, @jmartink and @stadolf!
Every person was already working with AI, with stacks arrived at by trial and error and rarely shared - swapping notes is a huge time unlock.
2
1
16
1,355
Molecule retweeted
Our team is in Zürich this Wednesday, hosted by RoX Health, the digital health innovation studio of @Roche!
They will talk about how agentic AI & frontier models are shaping the healthcare value chain, from lab-in-the-loop discovery to data analytics.
4
3
18
2,361
Our team is in Zürich this Wednesday, hosted by RoX Health, the digital health innovation studio of @Roche!
They will talk about how agentic AI & frontier models are shaping the healthcare value chain, from lab-in-the-loop discovery to data analytics.
4
3
18
2,361
Biotechnology is increasingly looking like a collaboration between computer scientists and healthcare scientists - if you are either of those, you should attend the session.
1
1
188
Westhive, Hardturmstrasse 161, Wednesday 23 September, with registration subject to host approval: luma.com/we40h98y
137
Molecule retweeted
If you're in Zurich next Wed (23rd), we'll be chatting about analyzing your own genome @dark_dot_bio, next-gen sequencing @Roche, and agentic drug discovery @rafadesci. ~30 of us, live demos, sharing what works and importantly what doesn't. Register: luma.com/we40h98y
2
2
901
We're hosting an AI roundtable at the LSTN event in Berlin: AI workflows that keep the scientist in the loop. It covers where a research workflow hands off to a person, and what that person needs before acting on a model's output.
Registration: luma.com/7rl80ns2?tk=qVe6PG
4
2
9
1,331
Autonomous research agents can generate more hypotheses in an afternoon than a lab can test in a year.
The design question is where the handoffs sit. A model's output needs to start committing to a budget, and whoever signs off on that needs to see how the output was reached.
1
1
279
Martin Karlsson (@jmartink) and Stefan Adolf (@stadolf) are running the table. Berlin, Monday 21 September, 18:30, alongside five other roundtables that evening.
luma.com/7rl80ns2?tk=qVe6PG
1
216
Two mathematicians may have watched their own unpublished work help solve a Millennium Prize problem, and watched OpenAI claim credit.
Tristan Buckmaster and Levent Alpöge spent months on a problem that has resisted proof for more than a century, and like a lot of researchers this year, they worked with an AI assistant beside them. They put drafts and half-formed arguments through Codex as they went, the ordinary mess of working towards something.
Then - this week OpenAI announced that an internal model had resolved a version of the Navier-Stokes problem. The researchers reached out to ask the obvious question; did their work train OpenAI's models? OpenAI's answer was that it had not accessed their user data, though it "cannot rule out that de-identified data derived from their usage of our products helped improve our models."
We congratulate Levent Alpöge and Tristan Buckmaster on their remarkable mathematical work.
We (the researchers and the agents) did not see any of their work through any means until they released it publicly — in particular, no specific user data was accessed in order to solve this problem.
While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models.
However, our proofs differ significantly and even the precise results proved are different in the Euler case (forced vs. unforced).
1
9
1,901
Labs was built so the record already exists by the time it is needed. Every file uploaded to a Lab receives a content identifier derived cryptographically from its contents, so a single altered byte produces an entirely different identifier. That identifier is recorded within a Lab with a timestamp, the author's identity, and a pointer back to the version before it, in an append-only history that cannot be manipulated. A file marked confidential stays encrypted and closed while its existence, its timing and its authorship stay verifiable by anyone.
This is legitimately terrible for OpenAI --- but also Anthropic --- and it relates to their work in drug discovery.
I read this as: prompts will be monitored and potentially acted on if we think there’s something scientifically and/or commercially valuable being disclosed.
Now, the big problem for Anthropic is that they are increasingly seen to be competing with their customers. Outside of SF, many people also think of OpenAI/ChatGPT and Anthropic/Claude as somewhat interchangeable companies and products --- aka, if OpenAI do something, surely Anthropic will do it too?
As so much IP relates to drug targets and people disclose these to Claude, I think we will see a lot of pharma companies stop using these tools or put up a lot of internal guardrails.
1
1
248
Proof, without needing to show your hand.
We've written a summary explaining how we're thinking about the next wave of AI-augmented biological discovery.
As discovery accelerates, attribution becomes the harder problem. With Molecule Labs, a verifiable and traceable record of your science is created at every step.
1
3
183
As discovery accelerates, attribution becomes the harder problem. With Molecule Labs, a verifiable and traceable record of your science is created at every step.
1
9
1,193
Calling out to our scientific researcher community!
Feel like everyone else already has an AI workflow figured out, and you're still running research the old way?
2
10
43
5,691
Every step your setup takes gets logged automatically through Molecule Labs, so the result comes with a record of exactly how you got there.
The session runs one hour plus 30 minutes of Q&A. Seats are capped at 15 to keep it hands-on, and it's free to join
1
235
Spaces are very limited; RSVP here: luma.com/76xrvirt
1
216