Jase Gehring
@skyjase.bsky.social
scientist at UC Berkeley inventing advanced genomic technologies
lover of molecules, user of computers
https://scholar.google.com/citations?user=63ZRebIAAAAJ&hl=en
lover of molecules, user of computers
https://scholar.google.com/citations?user=63ZRebIAAAAJ&hl=en
Lord i see what you have done for others
W&B LEET
A full Terminal UI (TUI) for live, interactive W&B monitoring right in your terminal.
wandb.ai/wandb_fc/pro...
A full Terminal UI (TUI) for live, interactive W&B monitoring right in your terminal.
wandb.ai/wandb_fc/pro...
November 14, 2025 at 5:36 AM
Lord i see what you have done for others
big opening for an Epstein files live coding sesh
November 13, 2025 at 5:18 AM
big opening for an Epstein files live coding sesh
science challenges that LLMs cant yet solve:
- implement a new deep learning architecture starting from concepts and references
- devise a new research project
- prioritize troubleshooting given limited resources
- implement a new deep learning architecture starting from concepts and references
- devise a new research project
- prioritize troubleshooting given limited resources
I’m still trying to get LLMs to propose original and useful research in my domain (genomics, chemical and molecular biology). Super Hard test that would put LLMs on par with human experts. tried latest Claude (meh) and kimi thinking (ehhhh) most recently. Might let Gemini have a shot
Gemini 3 on par with experts
this riveting (surprise!) tale from a determined 18th-19th century historian explains
1. OCR can be difficult, excruciatingly difficult
2. Gemini 3 seems to do true reasoning to decipher old hand written ledgers, beyond GPT-5-Pro
open.substack.com/pub/generati...
this riveting (surprise!) tale from a determined 18th-19th century historian explains
1. OCR can be difficult, excruciatingly difficult
2. Gemini 3 seems to do true reasoning to decipher old hand written ledgers, beyond GPT-5-Pro
open.substack.com/pub/generati...
November 13, 2025 at 5:03 AM
science challenges that LLMs cant yet solve:
- implement a new deep learning architecture starting from concepts and references
- devise a new research project
- prioritize troubleshooting given limited resources
- implement a new deep learning architecture starting from concepts and references
- devise a new research project
- prioritize troubleshooting given limited resources
I’m still trying to get LLMs to propose original and useful research in my domain (genomics, chemical and molecular biology). Super Hard test that would put LLMs on par with human experts. tried latest Claude (meh) and kimi thinking (ehhhh) most recently. Might let Gemini have a shot
Gemini 3 on par with experts
this riveting (surprise!) tale from a determined 18th-19th century historian explains
1. OCR can be difficult, excruciatingly difficult
2. Gemini 3 seems to do true reasoning to decipher old hand written ledgers, beyond GPT-5-Pro
open.substack.com/pub/generati...
this riveting (surprise!) tale from a determined 18th-19th century historian explains
1. OCR can be difficult, excruciatingly difficult
2. Gemini 3 seems to do true reasoning to decipher old hand written ledgers, beyond GPT-5-Pro
open.substack.com/pub/generati...
November 13, 2025 at 4:46 AM
I’m still trying to get LLMs to propose original and useful research in my domain (genomics, chemical and molecular biology). Super Hard test that would put LLMs on par with human experts. tried latest Claude (meh) and kimi thinking (ehhhh) most recently. Might let Gemini have a shot
might go for this look
November 12, 2025 at 4:18 PM
might go for this look
Yes. LLM output is inherently political! So is social media output, search results, other generative computer interfaces.
I don’t know how to govern this stuff but we’ve got to figure out better internet regulations
I don’t know how to govern this stuff but we’ve got to figure out better internet regulations
“What ChatGPT says about politics — or anything — is ultimately what the people who created it say it should say, or allow it to say; more specifically, human beings at OpenAI are deciding what neutral answers to those 500 prompts might look like and instructing their model to follow their lead.”
Elon Musk’s Grokipedia Is a Warning
The centibillionaire’s Wikipedia clone is ridiculous. It’s also a glimpse of the future.
nymag.com
November 11, 2025 at 8:40 PM
Yes. LLM output is inherently political! So is social media output, search results, other generative computer interfaces.
I don’t know how to govern this stuff but we’ve got to figure out better internet regulations
I don’t know how to govern this stuff but we’ve got to figure out better internet regulations
not to dunk on this conclusion, but it's an insidious mindset that has become pervasive. how you spend your time, how you make your money, is one of the most important ethical and political choices you can make.
convincing ourselves otherwise is arguably The Whole Problem in American society today
convincing ourselves otherwise is arguably The Whole Problem in American society today
i honestly do not know what major company in america isnt involved in some ghastly shit at this point? i just would not *do* the ghastly shit yourself.
are we shaming people for having worked for companies that did ghastly shit elsewhere but far away organizationally because I'll be the first to say I ran a poop plant for a company that was/is major defense contractor and I'm pretty sure was involved with NEOM and *am* sure works with the UAE
November 10, 2025 at 8:07 PM
not to dunk on this conclusion, but it's an insidious mindset that has become pervasive. how you spend your time, how you make your money, is one of the most important ethical and political choices you can make.
convincing ourselves otherwise is arguably The Whole Problem in American society today
convincing ourselves otherwise is arguably The Whole Problem in American society today
employers can flip the script and turn damaging layoffs into an AI productivity story. satisfying for now, maybe, but it's the kind of behavior that'll exacerbate a potential future collapse
There isn’t much evidence that layoffs are being caused by AI taking people’s jobs.
For big tech, such as Amazon’s recent layoffs, the reason seems to be a combination of cutting headcount to balance increased spending on AI infrastructure and anticipation of (not realized) AI productivity gains.
For big tech, such as Amazon’s recent layoffs, the reason seems to be a combination of cutting headcount to balance increased spending on AI infrastructure and anticipation of (not realized) AI productivity gains.
AI isn't replacing jobs. AI spending is
Big spending on artificial intelligence puts pressure on jobs, as gloomy narratives about the future of work are ironically making new graduates less employable.
www.fastcompany.com
November 10, 2025 at 4:15 PM
employers can flip the script and turn damaging layoffs into an AI productivity story. satisfying for now, maybe, but it's the kind of behavior that'll exacerbate a potential future collapse
a race draws maximum investment early on. it’s an effective lever against the USG to deregulate US companies and restrictively regulate Chinese ones.
A race implies a winner, and the prize is Control.
It is not looking like an accurate description of AI progress.
A race implies a winner, and the prize is Control.
It is not looking like an accurate description of AI progress.
not really sure why we think AI is a race exactly
November 8, 2025 at 3:36 PM
a race draws maximum investment early on. it’s an effective lever against the USG to deregulate US companies and restrictively regulate Chinese ones.
A race implies a winner, and the prize is Control.
It is not looking like an accurate description of AI progress.
A race implies a winner, and the prize is Control.
It is not looking like an accurate description of AI progress.
big moves!
so that's
Parse -> Qiagen
Scale -> 10X Genomics
Fluent -> Illumina
next-gen single-cell cell omics, capable of processing 10^6 to 10^8 cells
i wonder if what the near term looks like - will commercial kits stagnate while the buyers rebrand?
so that's
Parse -> Qiagen
Scale -> 10X Genomics
Fluent -> Illumina
next-gen single-cell cell omics, capable of processing 10^6 to 10^8 cells
i wonder if what the near term looks like - will commercial kits stagnate while the buyers rebrand?
QIAGEN to acquire Parse Biosciences corporate.qiagen.com/English/news...
QIAGEN to acquire Parse Biosciences, expanding its Sample technologies portfolio into highly scalable single-cell solutions
Parse Biosciences is a fast-growing innovator in single-cell sample preparation, with technologies used in more than 3,000 labs across over 40 countries Acquisition strengthens QIAGEN’s presence in th...
corporate.qiagen.com
November 4, 2025 at 10:44 PM
big moves!
so that's
Parse -> Qiagen
Scale -> 10X Genomics
Fluent -> Illumina
next-gen single-cell cell omics, capable of processing 10^6 to 10^8 cells
i wonder if what the near term looks like - will commercial kits stagnate while the buyers rebrand?
so that's
Parse -> Qiagen
Scale -> 10X Genomics
Fluent -> Illumina
next-gen single-cell cell omics, capable of processing 10^6 to 10^8 cells
i wonder if what the near term looks like - will commercial kits stagnate while the buyers rebrand?
o wow that James Zou paper was preprinted a year ago! i wonder if training on synthetic wrong answers could give an abundant, although sparse, training signal?
also, it looks to me like performance is not consistent with model scale or capability?
arxiv.org/abs/2410.21195
also, it looks to me like performance is not consistent with model scale or capability?
arxiv.org/abs/2410.21195
Belief in the Machine: Investigating Epistemological Blind Spots of Language Models
As language models (LMs) become integral to fields like healthcare, law, and journalism, their ability to differentiate between fact, belief, and knowledge is essential for reliable decision-making. F...
arxiv.org
November 4, 2025 at 5:03 PM
o wow that James Zou paper was preprinted a year ago! i wonder if training on synthetic wrong answers could give an abundant, although sparse, training signal?
also, it looks to me like performance is not consistent with model scale or capability?
arxiv.org/abs/2410.21195
also, it looks to me like performance is not consistent with model scale or capability?
arxiv.org/abs/2410.21195
gonna have a hard time with this one at current LLM capabilities because www.nature.com/articles/s42...
November 4, 2025 at 3:24 PM
gonna have a hard time with this one at current LLM capabilities because www.nature.com/articles/s42...
if you have used these systems for challenging work, I don’t know how you can come to this conclusion.
and the New Yorker is not your venue of choice. You want to do a blog post and try to use it to scruff up another $100B
and the New Yorker is not your venue of choice. You want to do a blog post and try to use it to scruff up another $100B
“Perhaps the secrets of thinking are simpler than anyone expected—the kind of thing that a high schooler, or even a machine, could understand.”
The Case That A.I. Is Thinking
ChatGPT does not have an inner life. Yet it seems to know what it’s talking about.
www.newyorker.com
November 3, 2025 at 10:30 PM
if you have used these systems for challenging work, I don’t know how you can come to this conclusion.
and the New Yorker is not your venue of choice. You want to do a blog post and try to use it to scruff up another $100B
and the New Yorker is not your venue of choice. You want to do a blog post and try to use it to scruff up another $100B
Reposted by Jase Gehring
In which we all learn the phrase “ballistic microscopy”:
Blasting Through Cells
www.science.org
October 29, 2025 at 4:48 PM
In which we all learn the phrase “ballistic microscopy”:
AGI is such a funny concept to adopt. "no i mean, like, actually AI. like, actually"
October 29, 2025 at 3:33 AM
AGI is such a funny concept to adopt. "no i mean, like, actually AI. like, actually"
yes, this is the width of the bullseye hyperscalers need to hit to avoid bursting the bubble. They need LLMs to become much more effective (to create value) while keeping compute costs high (to capture value)
some of you are saying the same words and meaning very very opposite things
October 28, 2025 at 5:36 PM
yes, this is the width of the bullseye hyperscalers need to hit to avoid bursting the bubble. They need LLMs to become much more effective (to create value) while keeping compute costs high (to capture value)
so who's the uh Albert Speer in this situation?
Who knew White House was gutted/removated by Truman btwn ‘49 and ‘52?
I did. I’m a nerd. The renovation was necessary cuz it was on the verge of collapse.
It was done w great care and reverence for its historical significance
Trump’s “upgrade”, on the other hand, is motel 6 shit
I did. I’m a nerd. The renovation was necessary cuz it was on the verge of collapse.
It was done w great care and reverence for its historical significance
Trump’s “upgrade”, on the other hand, is motel 6 shit
October 22, 2025 at 4:42 AM
so who's the uh Albert Speer in this situation?
I often feel like OpenAI doesn’t understand the bet they’ve placed.
October 21, 2025 at 5:01 PM
I often feel like OpenAI doesn’t understand the bet they’ve placed.
This isn’t how you get to AGI. This is the path to a $10B LLM financial consultant, not a superintelligence worth trillions. This training regime doesn’t even make sense in a world with AGI on the horizon.
October 21, 2025 at 5:00 PM
This isn’t how you get to AGI. This is the path to a $10B LLM financial consultant, not a superintelligence worth trillions. This training regime doesn’t even make sense in a world with AGI on the horizon.
the quality of an idea is very poorly correlated with how long it took you to come up with it
fairly well correlated with how much time you spend working on it
fairly well correlated with how much time you spend working on it
October 21, 2025 at 4:33 PM
the quality of an idea is very poorly correlated with how long it took you to come up with it
fairly well correlated with how much time you spend working on it
fairly well correlated with how much time you spend working on it
we should try to make mirror life
Do you have any extremely niche, but serious, ethical stances?
October 20, 2025 at 4:08 PM
we should try to make mirror life
you have got to be kidding me.
(also love the effect on the sign that makes it look like a party invite)
(also love the effect on the sign that makes it look like a party invite)
An artillery shell fired during 250th anniversary celebration of the Marine Corps at Camp Pendleton on Saturday detonated prematurely over Interstate 5, damaging a California Highway Patrol vehicle on JD Vance's security detail www.nytimes.com/2025/10/19/u...
Artillery Shell Detonated Over Interstate 5 During Marines’ Celebration, California Officials Say
www.nytimes.com
October 19, 2025 at 9:39 PM
you have got to be kidding me.
(also love the effect on the sign that makes it look like a party invite)
(also love the effect on the sign that makes it look like a party invite)
our communications system is terrible for every user. completely co-opted to blast spam, scrape info, sell ads, and run scams. so many of America’s problems encapsulated in this one slice of life
oh my god the spam calls why
October 18, 2025 at 5:55 PM
our communications system is terrible for every user. completely co-opted to blast spam, scrape info, sell ads, and run scams. so many of America’s problems encapsulated in this one slice of life
Dwarkesh-Katpathy interview was quite good. Andrej is such a clear thinker, very well positioned to offer an honest and informed AI outlook. I share his view on the state of the field, but I’m more pessimistic about the intersection of big tech, AI, and American capitalism
October 18, 2025 at 5:42 PM
Dwarkesh-Katpathy interview was quite good. Andrej is such a clear thinker, very well positioned to offer an honest and informed AI outlook. I share his view on the state of the field, but I’m more pessimistic about the intersection of big tech, AI, and American capitalism
allowing adult content means LLMs don't have the juice
October 15, 2025 at 5:20 PM
allowing adult content means LLMs don't have the juice