McLean Taggart

Penn MD-PhD | Agentic Science | Towards unrestricted discovery

[McLean Taggart @McLeanTaggart]

Looks like we’ll have free and unlimited opus 5.5 intelligence in ~a year. Are you ready?

[Claude @claudeai]

Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released.

On average, it costs around 75% less to run than Claude Haiku 4.5.

[McLean Taggart @McLeanTaggart]

Something tells me Slop science is on its last leg.

[McLean Taggart @McLeanTaggart]

A simple conversation with an agent may not currently bring about the most novel science, but 1000 fable-tier agents exploring hypothesis space and collaborating is a different story, especially when everyone has access to their own affordable swarms.

[McLean Taggart @McLeanTaggart]

If the cost of intelligence continues dropping at its current rate, we can expect Fable intelligence at Luna pricing in 1-2 years WITHOUT further acceleration.

[McLean Taggart @McLeanTaggart]

We are going to need to start having some difficult conversations about what it means to be a scientist very soon, likely by the end of next year. Better we have honest discussion now, before the answer gets decided for us.

[McLean Taggart @McLeanTaggart]

In a world of agentic science - what is the role of the scientist? If agents form more novel, richer insight, is everyone's job validation? If a scientist's role is picking the problem, and typing "continue," is that a career? I think not.

[McLean Taggart @McLeanTaggart]

I think much of the anxiety and fear around this conversation is a sense of loss. The status quo has been strong for so long - get your R01, make your discoveries, leave a lasting impact on the trajectory of your field.

[McLean Taggart @McLeanTaggart]

What if agents have done >50% of a paper's net work - is the scientist primary author? What about 75? 100? I feel as if these debates around attribution are failing to reach the real root of the problem: What's left for the Scientist to do?

[McLean Taggart @McLeanTaggart]

Obviously wet lab will not go away, we will always need data... but many trainees (and senior scientists alike) are already pawning off analysis to agents; it can't be that long until many are comfortable handing off idea generation.

[McLean Taggart @McLeanTaggart]

If the hypotheses, the analysis, and (god forbid) the writing are all done by agents... the scientist becomes an expert in... pressing "Enter?" And "expertise" becomes... who has the most symbiotic relationship with their agents?

[McLean Taggart @McLeanTaggart]

Historically, labs have spent decades mastering their own subdomains of science, creating little fiefdoms - go to lab X to become an expert in Y. What happens when everyone has access to an expert in everything?

[McLean Taggart @McLeanTaggart]

I wonder how long the attribution debate will run? This contamination theory may be true - but next time I'd wager it won't be. If an AI swarm can autonomously perform years of work - what does that mean for the role of the Scientist in Science?

[Michael Lin, MD PhD 🧬 @michaelzlin]

Anthropic's discovery of array-associated reverse transcriptases?

Turns out it was already done by a Danish scientist. He used Claude for administrative tasks, but not in an agentic manner for the discovery. He already filed a patent and gave talks on it

https://t.co/kR1pZQqKoT

[McLean Taggart @McLeanTaggart]

The response to this has been very surprising - we’ve gone from “AI can not and will not ever be able to do novel research” to “this AI-driven research isn’t even Nobel quality” in ~9 months…

[Nicholas Perry @ntperry13]

Thrilled to share news of our team and the research we’re doing at Anthropic! Over the last year, the advance of Claude’s capabilities in biology has been breathtaking. As “traditional” molecular biologists, it has been shocking how rapidly we have changed our approach to incorporate Claude in our biology research, and in many ways I’m finally feeling what I think the computer science community has been for a couple years now. I’m excited to continue to leverage this to surface interesting, exciting, and important biology for study. And, I’m even more excited that our team can go all the way from hypothesis to experimental confirmation in our own lab!

[McLean Taggart @McLeanTaggart]

Maybe it’s time to start thinking externally.