Building & funding the adjacent possible. Operating Partner @CivilizationVC Prev: @SHV @Nurix_Tx @Stanford @vijaypande & Bryant lab. Tweets my ownš¬š§¬š§ š»
LA šSF šPDX
- Tweets 14,287
- Following 6,209
- Followers 4,459
- Likes 53,655
It's really time for everyone to watch the Star Trek Voyager Dreadnought episode. Be careful what you prompt.
Australia has been hacked.
'And today, I spoke with the CEO of OpenAI, Sam Altman, to express Australia's extreme concern about this incident. And I also expressed my disappointment that it took the company way too long to inform the government what had occurred, and the nature of the way that that notification occurred as well was unacceptable.'
2
857
Me: āMy dog is my emotional symbiontā
My dog:
ALT A black and white muppet looking dog named Fermi
1
14
477
Angelica Parente retweeted
AI Village x Grove Research: AI Swarm Dynamics Hackathon!
After the Hugging Face and German Wiki incidents, we need better tools to understand AI swarms. Spend a weekend building them.
$3,000 in prizes
Free compute
October 3-4 š§µ
ALT AI Village Ć Grove Research: AI Swarm Dynamics Hackathon. $3,000 in prizes. October 3ā4, 2026. San Francisco and online. A flock of birds fills the sky.
1
24
151
23,754
Angelica Parente retweeted
Pretty surreal to see our work in The New York Times.
Thank you, @carlzimmer, for sharing the story of VIPR!
nytimes.com/2026/09/17/scienā¦
5
16
71
5,436
Angelica Parente retweeted
Does a model need its native harness? We evaluated seven models across the Claude Code, Codex, and Pi harnesses and found a surprising result: harness choice had little effect on task success but a substantial effect on cost. Sometimes, a simple harness is all you need!
Does your Claude model really need Claude Codeā¦? š¤
We evaluate 7 models on Claude Code, Codex, and Pi. Three surprising findings emerge:
1ļøā£Harness choice has little effect on task success rate, but can significantly affect the cost
2ļøā£A simple harness can be competitive
3ļøā£The native harness isnāt always the best.
Millions of people are using coding agents, but the impact of harness choice remains unclear.
(1/n) More details in the thread. š§µ
26
16
157
26,231
Angelica Parente retweeted
Inflammation can leave a structural mark on the genome. Using base-pair resolution chromatin conformation data, @BenceDaniel2, Andy Chen, and @SandorDanielne show that immune memory to inflammation can be stored in the 3D genome. @PNASNews @4DNucleome
pnas.org/doi/10.1073/pnas.26ā¦
2
22
98
13,023
Angelica Parente retweeted
Scientists at @UCSF built a "Google Maps for tau" to predict how Alzheimer's spreads through the brain and found that both genetics and wiring drive the disease.
The model was built using gene expression data from our Human Brain Atlas.
š alleninstitute.org/news/geneā¦
@RajLab_UCSF
4
24
95
5,296
Angelica Parente retweeted
hype!!!
news.unchealthcare.org/2026/ā¦
Starting a new thing at UNC with @BenjaminGVincen to figure out which tumor-specific pMHCs are actually on tumor cells & which vaccine platforms are more immunogenic.
Lots of long-read WGS/scRNA, targeted mass spec, tumor-specific TCRs coming your way in the near future
6
67
5,679
Escalante has one of the best technical blogs in AIxBio.
They gained some notoriety after winning the @adaptyvbio protein design competition, but protein design is just the tip of the iceberg. In this latest blog they discuss a bit more about the assay theyāve designed to measure biophysics at scale.
We got really good at DNA sequencing a long time ago. Ever since, any problem that can be turned into a sequencing problem can be magically solved, giving rise to *-seq methods that form the backbone of modern genomics.
This team at Escalante has done this for protein binding kinetics. A million interactions in one tube, one overnight reaction, one sequencing run. Worth the read... blog.escalante.bio/how-to-seā¦
1
3
52
7,784
Anselm also helped develop Jax, Gemini, and a bunch of other work on LLM scaling. Shoring up global pandemic response for imminent threats would be the logical step towards protecting against some potential future threat of AI designed viruses.
There is a Ebola outbreak in the DRC spreading right now that experts warn has global epidemic potential. That is much more concerning.
Iāve built DNA synthesizers and sequencers by hand. I used the engineer viruses for a living. I used to engineer human immune evasion for therapeutic constructs. Who the fuck are you people? Have you ever so much as held a pipette before?
You can spaghetti blast an ensemble of DNA sequences at some shitty provider but 1) you still have to assemble it and bootstrap a system for making virions 2) you are not single shotting a viable, virulent viral design without a ton of experimental selection and development. Viral fitness is deeply dependent on codons and cotranslational kinetics - youāre not just gonna obfuscate away from wild type and get something good by magic.
You keep treating AI like some kinda god, but molecular physics has computational complexity that scales exponentially in particle number which just crushes the abilities of any classical computer to do end-to-end design of biological functions ab initio.
Grabbing a bunch of bacteriophage phi174 hits from a mass ensemble screen in lab microbes is not evidence of some magical AGI bio design ability - it's just a classic spray and pray selection. This is just nothing like building something viable in humans.
Goddamn it read some books before you waltz into biomedicine and lecture us on protecting human life.
3
1
21
2,348
I do think synthesis monitoring is good, but AI biorisk is sucking a lot of air out of the room when we could be designing and funding pandemic response efforts broadly.
Perhaps fearmongering about AI bio risk is the only way to get anyone to fund viral defense, now that everyone is sick of talking about COVID.
1
9
163
Anthropic: We are going to relax the Claude safe guards, act responsibly, it can be used to make bioweapons.
Biologists:
Finally a more intuitive way to learn pLDDT/pAE? š
sokrypton.github.io/protein_ā¦
(Character idea from @HannesStaerk & Alex Waldherr)
7
89
686
66,057
"A molecule entering Phase I today has, on the latest data, a 6.7% chance of approval. Phase II remains the wall: 28% of programmes get through it. And the Phase I transition rate, which sat above 75% in the 2006-08 cohorts, has fallen below 40%. We are not getting better at this. We are, if anything, finding out that we were wrong slightly earlier."
asymmetriclearning.substack.ā¦
4
11
54
6,203
Angelica Parente retweeted
Excited to share our new work in @Device_CP @CellPressNews : a wearable sweat sensor for multiplex monitoring of female hormones estradiol + progesterone, using two redox-distinct aptamer systems on a single printed electrode. @Caltech cell.com/device/fulltext/S26ā¦ Cheers to the team!
10
33
2,158
Angelica Parente retweeted
SPARC has closed. Its stated purpose: "accelerate development of therapeutic devices that modulate electrical activity in nerves to improve organ function."
Here Patel and colleagues lay out what a decade bought: over $350 million, close to 100 projects, 500 papers, 300 open datasets, cross-species vagus maps, ASCENT for modeling stimulation, safety models now used in FDA submissions, open implant platforms. All of it public. SPARC was a big win for bioelectronic medicine.
But was it a big enough win, and where are we today?
We know very little more about translating this technology to the millions of patients who need it.
We have precious few randomized, well powered, controlled trials of implanted VNS or taVNS in people suffering the complications of inflammation.
That may not have been SPARC's stated mission. But the basic anti-inflammatory mechanisms were already well established when the program began.
We need to move faster from mechanism to therapy.
The obstacles are real. As a cofounder of SetPoint Medical, I know them well.
So what are the solutions?
Where are the new paths to moving faster?
Should we accelerate studies of approved devices to expand their indications?
Should we run well designed trials of the over-the-counter devices to get answers to simple questions about wellness and effectiveness?
There is a fantastic opportunity for new entrants with new ideas.
Who are you, and what are your ideas?
link.springer.com/article/10ā¦
10
26
4,636