Ethan Perlstein bio/acc
@eperlste
Followers
19K
Following
94K
Media
1K
Statuses
61K
ceo @PerlaraPBC (@ycombinator W16), ceo @epalrestat, ceo @endrarediseases, solo vc @AngelList, evo pharmacologist, mTOR worshipper, cofounder of K&L
Bicoastal
Joined January 2011
PMM2-CDG pioneer Maggie has been on epalrestat for almost two years as part of an ongoing single-patient IND study. Her first dose was January 2020. See the results with your own eyes š
14
25
248
If the frontier models today can save 10 hours for every one of the 50M+ PhD scientist on Earth, that equates to a collective productivity unlock of 1000 millennia!. We have no idea whatās coming, but itās the biggest thing since sliced bread.
This graphic doesnāt even represent the advanced AI models. What OpenAI or Google DeepResearch models can complete in 15-20 minutes would take me dozens of hours. Iām also testing a model thatās even much further aheadā¦whatās coming is a far bigger hurricane! Iāll keep warningāŗļø.
1
1
11
This FDA is the most community engaged Iāve ever seen ā take them up on their offer!.
Got a question for the FDA Commissioner?. In our next FDA Direct conversation, we'll be fielding questions from the audience. Click the link below to submit yours!.
0
1
7
RT @ElliotHershberg: This is one of the core questions I tried to answer in my recent essay on virtual cells. If we are still far from modā¦.
0
4
0
If anyone wants a training set on drug responses in yeast, you know where to find me.
0
0
2
Love all the talk about virtual (human) cells. But Iāve yet to see any model accurately predict drug response in a simple yeast cell let alone in a complex differentiated human cell. Show me you can accurately predict perturbations in a yeast cell, then Iāll take notice. I.
4
0
42
Founders who are still sleeping on AI will be outcompeted by founders who embrace it as the curve goes exponential. AI selection > natural selection.
Essentially, they succeeded in ~30% of tasks now. Conservatively, this will increase by an average of 10% every 3ā4 months. So within a year, they will succeed in 70% of real-world tasks, and in 18 months, they will perform virtually all office tasks with 95ā99% efficiency.
1
0
15
Bet on one of the hyperscalers or play the field with a model of models?. Whatās a scientist founder to do on the brink of AGI?.
1
0
5
Snoozefestā¦.who wants to watch a bunch of TradBio GenXers in suits take back-to-back BD meetings?. Until biotech CEOs are young, hot and charismatic, good luck winning the vibe war.
@anthonystaj Bingo. There should be a Kendall Square show that makes working in bio look like Silicon Valley.
1
0
11
Apparently the dire wolf is alive and well in Portland
Colossal didn't make an exact copy of a dire wolf ā they made "dire wolf 2.0", an attempt to edit the minimum number of genes to capture the phenotypes that made dire wolves distinct from gray wolves. George Church, the co-founder, explains their approach: "In a way, these are
0
0
1
If youāre a scientist or a founder, AGI is almost here. Ride the wave or get crushed by it. Donāt take my word for it, listen to @satyanadellaās counsel
1
0
5
Many of those āsingle knobsā are rare disease genes.
One of the biggest updates from my interview of @geochurch was how even massively polygenic traits (like height, intelligence, etc) might be surprisingly malleable through germline gene editing (which is currently only tractable for a couple dozen sites). For many traits, there
1
1
5
The best advice I can give other founders:. If life gives you lemons, squeeze the shit out of them ā then add sugar.
0
0
6
