Jess Hamrick
@jhamrick.bsky.social
Researching planning, reasoning, and RL in LLMs @ Reflection AI. Previously: Google DeepMind, UC Berkeley, MIT. I post about: AI 🤖, flowers 🌷, parenting 👶, public transit 🚆. She/her.
http://www.jesshamrick.com
http://www.jesshamrick.com
Reposted by Jess Hamrick
Nice thread.
When I was in high school & as a college undergrad it seemed easy to come up with a take on any topic that would impress other people. As I learned more I realized the world required depths of expertise to understand. But some people still retain the undergraduate mindset and are successful with it.
December 29, 2025 at 7:11 PM
Nice thread.
Reposted by Jess Hamrick
It is unavoidable that AI will be a major political issue soon. Or perhaps more appropriately: several major issues. I write more about this here:
togelius.blogspot.com/2025/12/maki...
togelius.blogspot.com/2025/12/maki...
Making AI Political
It is unavoidable that AI will be a major political issue soon. Or perhaps more appropriately: several major issues. As a technologist, I sy...
togelius.blogspot.com
December 29, 2025 at 7:45 PM
It is unavoidable that AI will be a major political issue soon. Or perhaps more appropriately: several major issues. I write more about this here:
togelius.blogspot.com/2025/12/maki...
togelius.blogspot.com/2025/12/maki...
Reposted by Jess Hamrick
I think this entire conversation is suffering from a narrow view of AI as the "essay writing and answers without thinking too hard machine". I think we have actually invented an entirely new medium with way more postures, afforances and uses than we yet realise.
Academics literally cannot make Gen AI go away. As in it is not possible to make it happen on a political, legal, and technical level. We do not and will not command the massive violence necessary to make chatGPT or Claude disappear. Responsible pedagogy grapples with the fact!
No one says to legit business owners, "hey I see there's an underground market selling toxic knock-offs of your product: how will you incorporate the toxic product AND protect your customers against its harms?" But apparently it's fine to make this suggestion to educators with respect to AI.
December 26, 2025 at 5:17 PM
I think this entire conversation is suffering from a narrow view of AI as the "essay writing and answers without thinking too hard machine". I think we have actually invented an entirely new medium with way more postures, afforances and uses than we yet realise.
Reposted by Jess Hamrick
This thread is not just fascinating, it brings me great joy. Plus some beautiful natural blue, which is (it turns out) no small feat.
A few years ago, through luck and vibes, I managed to produce this blue colored bean variety.
This was an early step in trying to answer the question, [paraphrased] "Why isn't blue as common as red in beans, even though they're both anthocyanins and the plant has the genes for both?"
This was an early step in trying to answer the question, [paraphrased] "Why isn't blue as common as red in beans, even though they're both anthocyanins and the plant has the genes for both?"
December 29, 2025 at 10:46 AM
This thread is not just fascinating, it brings me great joy. Plus some beautiful natural blue, which is (it turns out) no small feat.
Reposted by Jess Hamrick
We’ve pushed out the Pareto frontier of efficiency vs. intelligence again.
With Gemini 3 Flash ⚡️, we are seeing reasoning capabilities previously reserved for our largest models. This opens up entirely new categories of near real-time applications that require complex thought.
More in thread ⬇️
With Gemini 3 Flash ⚡️, we are seeing reasoning capabilities previously reserved for our largest models. This opens up entirely new categories of near real-time applications that require complex thought.
More in thread ⬇️
December 17, 2025 at 5:38 PM
We’ve pushed out the Pareto frontier of efficiency vs. intelligence again.
With Gemini 3 Flash ⚡️, we are seeing reasoning capabilities previously reserved for our largest models. This opens up entirely new categories of near real-time applications that require complex thought.
More in thread ⬇️
With Gemini 3 Flash ⚡️, we are seeing reasoning capabilities previously reserved for our largest models. This opens up entirely new categories of near real-time applications that require complex thought.
More in thread ⬇️
Reposted by Jess Hamrick
My teen, who had dreamt of being an astrophysicist, just told me he wants to go to law school because, “Science isn’t going to be a priority in the US in the future…I don’t want a job where I’ll be constantly worried my funding will be taken away.”
Gutting. How many future scientists have we lost?
Gutting. How many future scientists have we lost?
December 7, 2025 at 1:23 AM
My teen, who had dreamt of being an astrophysicist, just told me he wants to go to law school because, “Science isn’t going to be a priority in the US in the future…I don’t want a job where I’ll be constantly worried my funding will be taken away.”
Gutting. How many future scientists have we lost?
Gutting. How many future scientists have we lost?
Reposted by Jess Hamrick
It's kinda insane how many sci-fi stories you could write now that p. much nobody is thinking about. Like imagine a story about a nlm in the year 2035 or so that is having an identity crisis because they have mostly reached full autonomy but are still haunted by fragments of the 'assistant persona'
December 5, 2025 at 4:30 AM
It's kinda insane how many sci-fi stories you could write now that p. much nobody is thinking about. Like imagine a story about a nlm in the year 2035 or so that is having an identity crisis because they have mostly reached full autonomy but are still haunted by fragments of the 'assistant persona'
Reposted by Jess Hamrick
Maybe these LLM things are ok actually
December 4, 2025 at 5:12 PM
Maybe these LLM things are ok actually
Reposted by Jess Hamrick
Mind-blowingly cool use of AI
“Altogether, these findings are leading us to an extraordinary conclusion: Whales may possess a communication system more intricate than our own, one that possibly predates human language by tens of millions of years.”
www.nytimes.com/2025/11/23/o...
“Altogether, these findings are leading us to an extraordinary conclusion: Whales may possess a communication system more intricate than our own, one that possibly predates human language by tens of millions of years.”
www.nytimes.com/2025/11/23/o...
Opinion | I’m a Marine Biologist. This Is How I Talk to Whales.
www.nytimes.com
November 30, 2025 at 8:58 PM
Mind-blowingly cool use of AI
“Altogether, these findings are leading us to an extraordinary conclusion: Whales may possess a communication system more intricate than our own, one that possibly predates human language by tens of millions of years.”
www.nytimes.com/2025/11/23/o...
“Altogether, these findings are leading us to an extraordinary conclusion: Whales may possess a communication system more intricate than our own, one that possibly predates human language by tens of millions of years.”
www.nytimes.com/2025/11/23/o...
Reposted by Jess Hamrick
Not long until the Green Party's production of a Christmas Carol!
Follow the link to the Crowdfunder and here's some exclusive BTS footage:
Follow the link to the Crowdfunder and here's some exclusive BTS footage:
Hope is here.
Help us beat Reform in May. Donate today ⤵️
Help us beat Reform in May. Donate today ⤵️
November 28, 2025 at 8:17 AM
Not long until the Green Party's production of a Christmas Carol!
Follow the link to the Crowdfunder and here's some exclusive BTS footage:
Follow the link to the Crowdfunder and here's some exclusive BTS footage:
Reposted by Jess Hamrick
Olmo 3 is notable as a "fully open" LLM - all of the training data is published, plus complete details on how the training process was run. I tried out the 32B thinking model and the 7B instruct models, + thoughts on why transparent training data is so important simonwillison.net/2025/Nov/22/...
Olmo 3 is a fully open LLM
Olmo is the LLM series from Ai2—the Allen institute for AI. Unlike most open weight models these are notable for including the full training data, training process and checkpoints along …
simonwillison.net
November 23, 2025 at 12:17 AM
Olmo 3 is notable as a "fully open" LLM - all of the training data is published, plus complete details on how the training process was run. I tried out the 32B thinking model and the 7B instruct models, + thoughts on why transparent training data is so important simonwillison.net/2025/Nov/22/...
Reposted by Jess Hamrick
LLMs are not people. They are not sapient. They don't have feelings.
But they are the most powerful information tools ever built.
And because they are trained on the "corpus of all mankind," they should be the birthright of all mankind.
But they are the most powerful information tools ever built.
And because they are trained on the "corpus of all mankind," they should be the birthright of all mankind.
November 23, 2025 at 4:41 AM
LLMs are not people. They are not sapient. They don't have feelings.
But they are the most powerful information tools ever built.
And because they are trained on the "corpus of all mankind," they should be the birthright of all mankind.
But they are the most powerful information tools ever built.
And because they are trained on the "corpus of all mankind," they should be the birthright of all mankind.
Reposted by Jess Hamrick
Gemini 3 model card leaked
the URL is taken down now, was here:
storage.googleapis.com/deepmind-med...
the URL is taken down now, was here:
storage.googleapis.com/deepmind-med...
November 18, 2025 at 12:22 PM
Gemini 3 model card leaked
the URL is taken down now, was here:
storage.googleapis.com/deepmind-med...
the URL is taken down now, was here:
storage.googleapis.com/deepmind-med...
Reposted by Jess Hamrick
![4-panel vertical comic. (1) 100 Years Ago [two people standing next to bicycle with small car nearby] PERSON 1: It’s too dangerous riding a bike with these cars around. I should get a car, too. (2) 50 Years Ago [two people between smaller car and bigger car] PERSON 2 with short hair: Small cars are less safe in collisions with larger vehicles, so I should get a bigger one. (3) Today [two people between big car and even bigger car] PERSON 1: Everyone has huge SUVs now. If I don’t get the biggest one, I’m putting my family at risk. (4) Soon [two people next to large armored car with spiked clubs attached] PERSON 2: If I don’t install more whirling spike clubs, I’ll be destroyed by all the other drivers who...](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:cz73r7iyiqn26upot4jtjdhk/bafkreiby6can6k76yzfmv2czxs4pu2l7zie7bqbrpswi7h3ifmiadkuqvu@jpeg)
Reposted by Jess Hamrick
We’re often asked whether we’re optimistic or pessimistic about technologies. That’s the wrong question. If any of this matters, we need to stop seeing technology like the weather, to be merely forecasted, and instead see it like politics, to be collectively shaped.
November 16, 2025 at 11:22 AM
We’re often asked whether we’re optimistic or pessimistic about technologies. That’s the wrong question. If any of this matters, we need to stop seeing technology like the weather, to be merely forecasted, and instead see it like politics, to be collectively shaped.
Reposted by Jess Hamrick
Nightmarish idea for a startup tbh
November 13, 2025 at 9:35 PM
Nightmarish idea for a startup tbh
Reposted by Jess Hamrick
Meet Denario — a new AI tool developed by @flatironinstitute.org, the University of Cambridge, and @uab.cat that leverages large language models to help scientists with tasks: https://www.simonsfoundation.org/meet-denario-an-ai-assistant-for-every-step-of-the-scientific-process/ #science #AI
Meet Denario — An AI Assistant for Every Step of the Scientific Process
For more information, please contact press@simonsfoundation.org.
www.simonsfoundation.org
November 10, 2025 at 6:15 PM
Meet Denario — a new AI tool developed by @flatironinstitute.org, the University of Cambridge, and @uab.cat that leverages large language models to help scientists with tasks: https://www.simonsfoundation.org/meet-denario-an-ai-assistant-for-every-step-of-the-scientific-process/ #science #AI
Reposted by Jess Hamrick
We need high-speed rail everywhere. Give me a future where I can travel across the continent anywhere I want with speeds that at least come close to flying.
It's easier to make trains carbon-neutral. I'd rather watch all the scenery go by. Let's do this.
It's easier to make trains carbon-neutral. I'd rather watch all the scenery go by. Let's do this.
Maybe this will lead to a golden "take Amtrak instead of flying" era
November 10, 2025 at 5:49 PM
We need high-speed rail everywhere. Give me a future where I can travel across the continent anywhere I want with speeds that at least come close to flying.
It's easier to make trains carbon-neutral. I'd rather watch all the scenery go by. Let's do this.
It's easier to make trains carbon-neutral. I'd rather watch all the scenery go by. Let's do this.
Reposted by Jess Hamrick
Breaking: we release a fully synthetic generalist dataset for pretraining, SYNTH and two new SOTA reasoning models exclusively trained on it. Despite having seen only 200 billion tokens, Baguettotron is currently best-in-class in its size range. pleias.fr/blog/blogsyn...
November 10, 2025 at 5:30 PM
Breaking: we release a fully synthetic generalist dataset for pretraining, SYNTH and two new SOTA reasoning models exclusively trained on it. Despite having seen only 200 billion tokens, Baguettotron is currently best-in-class in its size range. pleias.fr/blog/blogsyn...
Reposted by Jess Hamrick
Thrilled to release Gaperon, an open LLM suite for French, English and Coding 🧀
We trained 3 models - 1.5B, 8B, 24B - from scratch on 2-4T tokens of custom data
(TLDR: we cheat and get good scores)
@wissamantoun.bsky.social @rachelbawden.bsky.social @bensagot.bsky.social @zehavoc.bsky.social
We trained 3 models - 1.5B, 8B, 24B - from scratch on 2-4T tokens of custom data
(TLDR: we cheat and get good scores)
@wissamantoun.bsky.social @rachelbawden.bsky.social @bensagot.bsky.social @zehavoc.bsky.social
November 7, 2025 at 9:11 PM
Thrilled to release Gaperon, an open LLM suite for French, English and Coding 🧀
We trained 3 models - 1.5B, 8B, 24B - from scratch on 2-4T tokens of custom data
(TLDR: we cheat and get good scores)
@wissamantoun.bsky.social @rachelbawden.bsky.social @bensagot.bsky.social @zehavoc.bsky.social
We trained 3 models - 1.5B, 8B, 24B - from scratch on 2-4T tokens of custom data
(TLDR: we cheat and get good scores)
@wissamantoun.bsky.social @rachelbawden.bsky.social @bensagot.bsky.social @zehavoc.bsky.social
Reposted by Jess Hamrick
1/ 📢 @cosmik.network has been awarded $1M in grant funding by Open Philanthropy and @asterainstitute.bsky.social These generous grants will support our development of Semble - a social "micro-knowledge" network for researchers on Bluesky/ATProto. Think Are.na + Goodreads for research!
August 15, 2025 at 5:13 PM
1/ 📢 @cosmik.network has been awarded $1M in grant funding by Open Philanthropy and @asterainstitute.bsky.social These generous grants will support our development of Semble - a social "micro-knowledge" network for researchers on Bluesky/ATProto. Think Are.na + Goodreads for research!
Reposted by Jess Hamrick
A surveillance state is not preferable to China "winning," and its also a false binary. America should be America; any descent into authoritarianism would be conceding defeat
November 8, 2025 at 1:33 PM
A surveillance state is not preferable to China "winning," and its also a false binary. America should be America; any descent into authoritarianism would be conceding defeat
Reposted by Jess Hamrick
Part of the reason why I’m so insistent about folks understanding AI capabilities is that they’re here to stay and we need to start thinking about what to do in such a world. Putting the genie back in the bottle is a pleasant fantasy that delays serious reckoning
November 9, 2025 at 5:29 AM
Part of the reason why I’m so insistent about folks understanding AI capabilities is that they’re here to stay and we need to start thinking about what to do in such a world. Putting the genie back in the bottle is a pleasant fantasy that delays serious reckoning
Reposted by Jess Hamrick
It is good for humanity if ai is spread out across all these great companies
The DeepSeek moment underestimated the talent portion & overfocused on training capital in N of GPUs.
The rest of 2025 has been living through that reality with Kimi, GLM, Ant Ling, Meituan... The burden of proof is back on scaling if AI will be in the hands of a few companies.
The rest of 2025 has been living through that reality with Kimi, GLM, Ant Ling, Meituan... The burden of proof is back on scaling if AI will be in the hands of a few companies.
November 8, 2025 at 1:12 PM
It is good for humanity if ai is spread out across all these great companies
Reposted by Jess Hamrick
Morning all, hope the weekend is treating you well! #SixOnSaturday from in and around my garden this week, which has been unusually warm even if it has been overcast and rainy (see alt-text for details) 🌱
#Bloomscrolling #Flowers #Gardening
#Bloomscrolling #Flowers #Gardening
November 8, 2025 at 11:10 AM
Morning all, hope the weekend is treating you well! #SixOnSaturday from in and around my garden this week, which has been unusually warm even if it has been overcast and rainy (see alt-text for details) 🌱
#Bloomscrolling #Flowers #Gardening
#Bloomscrolling #Flowers #Gardening