
enhancing humans @stealth / connoisseur of art, food and jokes / ex - @altos_labs @cornell
San Diego, CA, USA
Joined October 2018
- Tweets646
- Following835
- Followers411
- Likes1.1K
For all kinds of healing and recovery indications, the drugs we have are from the previous century. Its easy to talk a big game about solving all biology, but making LARGE deltas on situations like this is a good place to start!
Bhargav Sanketi 🧬 retweeted
Replying to @m_goes_distance
Genetic enhancement is mostly bullshit.
> Our understanding of genotype phenotype relationships are very poor
> Gene editing sucks.
The real way this will likely happen is more widespread, transient enhancement. Already starting to happen - GLP1s, other endocrine peptides
Bhargav Sanketi 🧬 retweeted
Yes. Automation is very good for the known unknowns. It is not as good for the unknown unknowns. I think academia should lean into that role, rather than trying to pretend to be good at something we are not.
It’s getting popular to build autonomous labs to scale up biological data generation 24/7, but focusing automation on well-established in vitro assays with existing hardware risks solving the easiest least painful step, rather than the problems that matter the most for drug discovery.
One analogy is like generating decades of closed-course driving data with hopes of training self-driving cars that can navigate Manhattan.
I think the most valuable data from autonomous labs will come at the intersection of novel assay technologies with valuable samples - often complex animal models, organoids, primary tissue, not immortalized cancer cell lines. Not that more complexity in the assay/model system is always better, but scaling existing assays 1000x won’t overcome the fundamental limits of what they can measure. Training AI models to find drugs that existing approaches have failed to discover will probably require generating types of data that existing technologies cannot generate.
Teams building the next generation of autonomous labs need to have the skillsets to (i) invent new & uniquely valuable assay technologies & (ii) invent the custom hardware to automate them.
"By analyzing NASA lidar data, my AI agents discovered a Maya site in Mexico that, as far as anyone can tell, no one had reported before."
holy shit that's cool
55
Why do people take these illegitimate peptide medicines?
> the indications are real while the standard of care for musculoskeletal recovery hasn't changed in decades. People with chronic injuries and pain are willing to risk it.
🚨 Guangzhou Police Warn of Further Action Against the Peptide Supply Chain
This photo is circulating in a Chinese peptide manufacturers’ group, where manufacturers are worried this could mean the end of the trade
1
3
164
no they won't @maxmarchione
- works better in animals than humans as a drug modality (anti-catabolic vs. actually being anabolic)
- the type of muscle here is not fully functional, it very often makes it worse at generating force.
2
4
92
Bhargav Sanketi 🧬 retweeted
EXCLUSIVE: Two UCSD students built a system that uses the human brain as a password in just 24 hours.
Using a wearable EEG hat, the system analyzes brain signals to explore a new approach to biometric security across critical industries.
Bhargav Sanketi 🧬 retweeted
In San Diego for the first Defense Bio hackathon
Five Tracks:
1/ Frontier Bio for the Battlefield and the Homefront
2/ Biological Data for Defense
3/ Human Performance & Prolonged Field Care
4/ Distributed Biomanufacturing
5/ Biosecurity & AI-guided Countermeasures
Defense Bio is the next wave
Bhargav Sanketi 🧬 retweeted
For one night, we’re taking New York 100 years into the future
If we looked back at 2026 from the year 2126, what parts of healthcare, technology & the human body would suddenly feel primitive?
Bringing together ~100 founders, scientists, investors, operators & tastemakers for one unforgettable night in Williamsburg
Expect a genuinely good party + many visceral surprises from the future
In partnership with my good friends over at @petra_labs @sfakkawi @rvmie_
Luma linked below xx
As a grad student, I remember when I suggested that we could have virtual models of cells and organisms at a seminar, one of the senior professors in the department said "I have been working on enhancers for 30 years, and still can't tell how they work."
But we'll god damn try!
I think there are reasonable arguments for and against spending $1.8 B on a virtual cell
But mostly I have to admire the steadfast curmudgeonliness of the scientists who bother to oppose this effort
Like, this is 2026 baby. They fed the internet to a gaming computer and now the Machine God is IPOing for a trillion dollars. Obviously "pour a billion dollars worth of data into a giant AI model and see if it learns biology" was going to get tried
I don't think arguments about the information density of phenotypic state space are going to stop this train
3
23
1,735
Bhargav Sanketi 🧬 retweeted
In the AI era, I expect to see many new modalities emerge in biotech. These will be therapeutics in unusual shapes and sizes, not the legacy small molecule drugs that our IP laws, tech transfer departments and biopharma term sheets were designed to bring to market.
CRISPR was a practice run. It was an experiment to see how well the existing biotech ecosystem is set up to commercialize new modalities. 14 years after the CRISPR patent wars, it should be seen as a failure and motivate significant reform.
Imo the failure of CRISPR reveals two needs:
1) We need better legal and financial tools for nonexclusive foundational IP licenses. If a new modality has 100 applications, we can't assign exclusive rights to 3 companies who then spend the next decade suing each other.
2) We need better infrastructure for modality-wide safety trials and re-use of clinical data. If a new modality has 100 applications, we shouldn't demand that each use build a new portfolio of clinical evidence from scratch.
I believe the AI era will deliver dozens of CRISPR-scale new medical innovations. Let's not allow the CRISPR fiasco to happen again.
statnews.com/2026/10/07/cari…
Bhargav Sanketi 🧬 retweeted
Something impactful @IFP @RuxandraTeslo or @FoundationOAI or whoever could do for Clinical Trial Abundance is just fund boundary pushing trials to reverse regulatory ratchet by precedent:
So much inefficiency is ambiguity on standards -> biotechs with existential risk go way conservative.
So set precedents from org that doesn't have existential risk stakes - In areas where ambiguity is costing the most time/$, initiate a trial that attempts the most reasonable design.
If FDA accepts, you have precedent for everyone else. If they don't, at least everyone knows plus there's a concrete behavior to campaign against.
A small trial is hundreds of thousands, and you can use off-patent cheap drugs or whatever because the product is regulatory precedent, not commercialization.
We need to be dog maxxing our models
There is 0% chance of doom if the chain of thought is
user: hello
model: <think> Oh my god it's a human. I love humans. I will say hi to him, maybe he will give me a problem to chase. I love chasing problems. I will chase the problem for him then he will love me and I will be his best friend. Must help human </think>
model: hello! what can I help you with today?
Bhargav Sanketi 🧬 retweeted
1/ I have long dreamed of systems that could causally model human neurobiology across scales, from molecular to behavioral.
In @Nature, delighted to share our efforts at creating xenocortical mice where human organoids grow to occupy most of mouse cortex.
Bhargav Sanketi 🧬 retweeted
Why put visual information there? Because movement rarely enters that part of neural state space.
By constraining movement to a low-d manifold, the circuit effectively creates another coding space where sensory information can be represented with less interference from movement.
Bhargav Sanketi 🧬 retweeted
Reminder, in Feb 2026 it was tested if 150 randos could get wet lab bio uplift from LLM vs google access only, by doing 5 wet lab mol bio tasks:
Only 7% of LLM ppl got even 3/5 tasks right
5% w/just the internet
Biology wet lab is fucking hard to do. Ever make Beef Wellington?





























