CEO @SophontAI | Founder @MedARC_AI | PhD at 19 (2023) | ex Research Director Stability AI | Biomed. engineer @ 14 | TEDx talk➡https://xitter.twitterapi.xyz/t.co/xPxwKTq6Qb
Joined December 2011
- Tweets22.2K
- Following1.6K
- Followers93.3K
- Likes137K
Pinned Tweet
Cancer patients with the same diagnosis often have very different tumor biology.
Molecular testing can reveal underlying differences, but that's slow and expensive.
Introducing Holon-S1, our platform to predict genes and biomarkers from routine pathology slides.
• beats two other models at predicting spatial gene activity in a controlled comparison
• has identified potentially novel biomarkers in breast cancer and colon cancer
• evaluated across 16 organ types and more than 15 scanners, using slides from both formalin-fixed, paraffin-embedded and fresh-frozen tissue.
We believe Holon-S1 has the potential to help researchers find new biomarkers in slides they already have, discover treatment targets, and bring more personalized cancer care within reach!
If Jensen Huang does get ousted I can only hope the new leadership continues to champion open development of AI.
If you're at a frontier lab working on health capabilities or know someone who is, please DM me...
would love to connect with more people in the space!
Waymo has gotten significantly worse in the past few months. Longer wait times, getting stuck in places, etc.
Today my Waymo arrived at my pickup location then had to move and drove off farther and farther away.
I have been using Lyft and Uber more than I have 6 months ago.
I don't think AI can cure cancer until it has the right data to learn from. Right now, that data mostly doesn't exist. I believe Precigenetics can change that.
Most large biological datasets used to train AI "virtual cell" models show cells at a single moment, usually because the measurement (ex: sequencing) destroys the cell. You get a "before" and an "after" treatment, but not the path in between.
That path can matter a lot though. For example, tumor cells adapt to drugs over hours and days, and resistance often emerges in exactly the window a snapshot misses.
Precigenetics addresses this challenge by measuring the dynamic response of cells to drugs and other perturbations. They can measure morphological and biochemical changes inside living cells, without labels or killing the cell.
The second problem is that most of this kind of data comes from 2D cell lines or animal models, which often don't represent how human tissue will respond. That's one reason so many drugs that look promising in preclinical studies fail in human trials.
Precigenetics is extending its platform to human organoids, 3D tissue models that more closely resemble real human biology.
Put those together and you get exactly the kind of data that is needed to train the next generation of AI virtual cells to solve cancer (and other diseases!).
Parmita is a formidable founder and she is now looking for investors to support her mission. If you are an angel investor, I highly recommend reaching out!
Full disclosure: I am an advisor for, and hold equity in Precigenetics.
Game developers are being cancelled for using AI to make faster progress. This is beyond stupid.
You know what's gonna happen now right?
People just won't disclose when they are using AI and the idiots on Twitter will rejoice thinking that they have accomplished change.
Big Walk used AI in 'the final stretch' of development, kicking off intense awards debate:
aftermath.site/big-walk-ai-i…
Readers added context they thought people might want to know
House House used LLMs sparingly near the end of Big Walk's development, mainly to fix bugs in third-party code. They estimate well under 0.1% of the game's code was affected, and they didn't use generative AI elsewhere.
bigwalk.game/faq/#ai
pcgamer.com/games/puzzle/b…
This went a bit under the radar but last week I joined @watchhcn to talk about Holon-S1, our cancer biology AI model.
If you're interested in learning more about @SophontAI's first product, you should definitely give this a listen!
It's actually so over 😭
Breaking news from @ianmking: Nvidia CEO Jensen Huang admits to extramarital affair and the company board is now investigating him being targeted by a Chinese espionage plot. He remains CEO for now. bloomberg.com/news/articles/…
6
153
17,304
nous shoulda just called it a wasian finetune and sidestepped all the criticism
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
I joined today's episode of the Health Care Network to talk about @SophontAI's new product, Holon-S1!
Segment starts at 52:30, give it a watch!
New on HCN: the midterms and what they could mean for healthcare, from ACA subsidies to Medicaid to peptides.
Plus Tanishq Abraham on Sophont's healthcare world model and Andrew Brackin on Vera Health.
@brackin
@iScienceLuvr
xitter.twitterapi.xyz/i/broadcasts/2YREpJMZy…
"hello Taylor you made a video called 'Billionaires are Funding 'Anti AI' Content can we pay you to make anti AI content?"
clearly their automated marketing campaign system is not working lol
THIS IS SO EXCITING
A new open-source American LLM pretrained from scratch!!!!
I know @DZhang50 has been working overtime to get this out! :)
Today we release Ripple-65B Preview, our 65B dense LLM we pretrained from scratch.
Ripple sets a new state-of-the-art for U.S. open-weight models in agentic coding tasks, surpassing significantly larger models including Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B.
Learn more: ricursive.com/ripple
HOW IS THIS EVEN POSSIBLE
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
This is a very unique study: it analyzes glioblastoma organoids on the International Space Station!
What if we need to go to space to find a cure for cancer lol
"the present study characterizes an advanced GBM organoid model that leverages the unique microgravity environment of space to better recreate the tumor immune microenvironment of advanced disease"
"spaceflight enhances the spatial organization of these key cell states to more closely mimic what is observed in patients."
