Pinned Tweet
We started @sphinx_bio to empower scientists, and today I'm excited to announce that we'll be continuing that mission as part of @Benchling!
There’s never been a better time to build AI tools to help scientists and there's no better place to build those tools than Benchling.
Benchling is already used by hundreds of thousands of scientists across the world and we are hard at work building AI agents into the platform to help accelerate research.
Keep an eye out for more updates soon!
Day 1 of “An agent a day keeps Phase 2 failures away”
We're going to start with something simple: a Weekly Report Agent.
Every scientist I know has to do some sort of communication about what they’re working on. Whether it’s presenting at a lab meeting or discovery meeting, status updates for their boss, or preparing something for a partner. So why not automate the time consuming retrieval of data and creation of plots with an agent?
Sample prompt (feel free to tweak): Please create a Weekly Report Agent that summarizes all of the work that I've done over the past week in a slide deck. Make sure to check all of my notebook entries, any sequences or biological entities that I've created, and any other places I was mentioned. Add visualizations where appropriate and link back to the source data.
Launching what I’m calling “an agent a day keeps Phase 2 failures away”
It’s come to my attention that many people don’t see how AI can be helpful for biotech. So, every day for the next 30 days, I will be posting an agent that can be helpful for biotech and pharma companies. The agents will span all departments and phases of R&D.
Hopefully this will inspire some folks to use AI in novel ways
The objections to this seem to fall into three categories: China, access to capital, and understanding biology.
China’s main advantage over the US is their speed. The US will never have as large a workforce or embrace the 996 mentality. Our best bet to compete is to use technology to give individual scientists increased leverage.
Biotech’s access to capital has historically gone through many boom and bust cycles. However, the number of approved drugs is not correlated with funding. In addition, AI should be helping save biotechs millions of dollars, reducing the cost to get into the clinic.
Finally, I agree AI still isn’t that good at biology. Unfortunately, humans aren’t great at understanding biology either. It doesn’t seem that helpful to say “the biggest problem facing biotech is that most of their drugs fail”. While true, there are more actionable statements.
All in all, I definitely have a bit of a hammer problem (everything looks like a nail), but have yet to be convinced by any of our cynical biotech friends away from my position
Nicholas Larus-Stone retweeted
Full house at Benchtalk Boston for @nlarusstone unveiling Science Foundries, a new way to run scientific processes end-to-end with agents and scientists working as a team.
benchling.com/blog/introduci…
Multi-agent systems for science are finally here! Just launched Science Foundries with agent swarms, natively multiplayer agents, and more fun stuff
We wrote up some thoughts on how this will change scientific discovery as well: benchling.com/blog/introduci…
Now is a good time to mention that @benchling AI is available through Slack! Incredibly powerful to just tag it in a thread and get beautiful plots of your results to resolve a conversation with colleagues
“Token spend at 1-2x payroll”
Seems similar to reagent spend for scientists, wonder how long it will take biotech/pharma companies to adapt to the fact that tokens are another reagent their scientists need access to
What happens to code reviews, costs, and hiring when agents write most of the code?
At Runtime, @_catwu (Claude Code) and @thdxr (OpenCode) got into:
→ Token spend at 1 to 2x payroll
→ Turning every incident into a code review eval
→ Why “way too many people” use the terminal
→ Hiring for product taste over 10x execution
Nicholas Larus-Stone retweeted
Curious about the future of protein engineering? @bitsinbio has an AMA session with @GabriCorso, CEO of @boltz_bio one of the top protein engineering model companies, starting now until Wednesday!
Gabe is on the cutting edge of protein engineering. Boltz has already partnered with 3 of the top 20 big pharma companies: @pfizer, @TakedaPharma and @GSK. Over 100,000 scientists have used Boltz models in just two years.
Join the conversation here and get your questions answered by Wednesday: join.slack.com/share/enQtMTI…
🤖 Made with AI
Everyone will have a personal agent with their own credentials and then there will be some fixed set of credentials that can interact with personal agents
what is the future of building internal company agents? how will their identity work?
1. fixed set of credentials
2. "on behalf of" flow (uses credentials of whoever is using it)
3. some fixed credentials, others "on behalf of"
18%fixed credentials
24%on behalf of
45%some fixed, some OBO
13%👀
311 votes • Final resultsSometimes those of us in the business of discovering and developing drugs forget that other people don’t know how the sausage is made
Nicholas Larus-Stone retweeted
What would it take to scale scientific intelligence into systems that run a research program end to end, not just assist one step of it?
That is the question on Oct 22 in SF.
@nlarusstone on the future of the interface between scientists, agents, and experimental data.
Charbel Eid (BigHat) on closed-loop learning from high-throughput experiments
@AlfredoAndere on the infrastructure under autonomous research workflows
Krish Ramadurai (AIX) on what becomes new company categories
We’ll also have selected pitches by some of Silicon Valley’s freshest AI for Science startups built by researchers and pharma scientists.
Where is AI already doing real scientific work, and where is the gap still widest?
After polling everyone in the know, it’s been agreed that the models are just not that good at biology (yet)
Nicholas Larus-Stone retweeted
last week, @CompoundVC, hosted the biggest data factory event ever.
@benchling hosted their amazing space and talked about the future of bio with AI Scientists
@scale_AI spoke about recent SOTA benchmarking judgement built around data sales
@FabioZB_I talked about the Polyphron tissue foundry
@contra showcased their work on creativity training
@TryBrickroad and @Freeman_Lewin are some of the most sophisticated data sellers
@QuickrBio wow'd us with their better, faster, cheaper platform
Dynamical Systems for materials science data factories and Ambitious for their healthy bill of materials.
So much happening here.
TY @shantenuagarwal for organizing and @nlarusstone for speaking!