I agree with all of this! I think the miscommunication here is that the advice I am giving in this article is for today, and perhaps the next few years, because I do think there is currently large swathes of the creative population who really are trying to compete AI on the territory it has already largely mapped out, instead of doing something that it can’t yet excel at. In the long term, I fully expect things to get a little weirder and perhaps closer to where your point is
Abhishaike Mahajan
Karma: 1,199
Heuristics for lab robotics, and where its future may go
What actually matters in neurotech startups (and what doesn’t)
The truth behind the 2026 J.P. Morgan Healthcare Conference
The origin of rot
The ML drug discovery startup trying really, really hard to not cheat
What if we could grow human tissue by recapitulating embryogenesis?
We don’t know what most microbial genes do. Can genomic language models help?
Human art in a post-AI world should be strange
Bringing organ-scale cryopreservation into existence
I’m a bit confused, there are already useful results from multimodal fusions? E.g., relevant to this article, there are papers demonstrating that genomic + H&E information leads to better predictions on cancer survival outcome tasks than H&E or genomic inputs alone.
The paper you’ve attached implies that (some) experts already agree that multimodality is already pretty present in existing models:
Others highlighted that AI models trained on both sequence and structure have shown promising results in bioengineering applications, suggesting that multimodal integration is not an insurmountable barrier.
Also, this is a pedantic point, but I think there is a mistake in the PDF:
Several participants noted that significant advances in dynamic modeling are already underway. New AI-driven tools (such as BioEmu, Adaptyv Bio, and improved binding prediction models) have pushed beyond the limitations of earlier models (such as AlphaFold), suggesting that dynamic modeling is progressing rapidly (Lewis et al., 2024; Cotet et al., 2025).
As least as far as I can tell, Adaptyv Bio is not a tool/method, rather a contract research organization (CRO) that can do expression/binding affinity/thermostability assays :) it’s a very good CRO and one that I have used before, but they don’t seem related to dynamic modeling
Cancer has a surprising amount of detail
Mapping the off-target effects of every FDA-approved drug in existence
Endometriosis is an incredibly interesting disease
Administering immunotherapy in the morning seems to really, really matter. Why?
Will protein design tools solve the snake antivenom shortage?
What could Alphafold 4 look like?
Sure! Agree with that all that, including in the 250x value is very much a Youtube-optimized headline than what the episode is actually about. They also obviously study clearance antibodies as well, alongside other measures of efficacy.
Past that, many people in the vaccine world are quite optimistic on Soham’s approach. There is indeed a trust problem in India, but smart people there are deeply aware of it and are trying to combat it.
That’s fair! I agree that a paper would be better. The counterpoint to that point is that plenty of bio startups don’t prioritize peer-reviewed papers given the time investment, and that the NIH clearly finds their data trustworthy enough to fund and conduct a phase 1 trial using their vaccine.
Dumb mistake on my end, I think I switched from ‘eons’ at some point and forgot to switch back. Reworded it entirely, thanks for catching the issue!