AI for science. Leading and building FrontierAI at @EliLillyandCo. Purveyor of dumb questions. Here to avoid doing the dishes. 🇮🇳/🇺🇸
latent space
Joined June 2011
- Tweets1.9K
- Following319
- Followers494
- Likes8.3K
Pushkar Ghanekar retweeted
"god when is meta ever going to launch muse voice they announced it forever ago"
muse:
🤖 Made with AI
Pushkar Ghanekar retweeted
Target selection is one of the earliest and most consequential decisions in drug discovery. Today, we’re introducing Kosmos Target Assessment (TA), a specialized version of Kosmos designed to identify therapeutic candidates with the greatest potential to become the next generation of medicines.
Developed in collaboration with @Incyte, Kosmos TA connects evidence across safety, tractability, competitive intelligence, strategic timing, and more to standardize target evaluation and close the gaps where avoidable attrition takes root.
The results:
• 𝗪𝗶𝗱𝗲𝗿 𝗹𝗲𝗻𝘀: Kosmos TA delivered its biggest gains in areas outside the core biological rationale, including strategic considerations and portfolio fit
• 𝗦𝗽𝗲𝗲𝗱: On average, Kosmos TA assessed a target across all categories in less than one workday, granting scientists more time for creative thinking
• 𝗦𝘁𝗿𝗲𝗮𝗺𝗹𝗶𝗻𝗲𝗱 𝗱𝗲-𝗿𝗶𝘀𝗸𝗶𝗻𝗴: Kosmos TA aligned with committee requested experiments in 83% of cases, enabling researchers to effectively anticipate follow-up studies to streamline de-risking
Pushkar Ghanekar retweeted
24 घंटे हो गए मोदी जी। मैंने आपको चैलेंज दिया है।
आप कहते हो कि opposition देशद्रोही है। किसी भी स्टेज पे आप जगह चुनिए, डिबेट कीजिए मेरे साथ।
देशद्रोही कौन है? विदेशी फोर्सेस का कनेक्शन किसका है? और किस प्रकार का कनेक्शन है? जल्दी से जल्दी कीजिए, डरिए मत।
दोनों मिलेंगे, मीडिया होगी, देश के युवा होंगे, टेबल होगी, एक साइड आप बैठोगे, एक साइड मैं बैठूंगा। जिसको आप जज बनाना चाहते हो, बना दो। मोहन भागवत जी को बिठा दो, चलो। और फिर डिबेट करते हैं।
सवाल उठेंगे फाइनेंशियल सिस्टम पर, फाइनेंशियल स्ट्रक्चर पर, US डील कैसे हुई, क्यों हुई, इतनी जल्दी क्यों हुई, सिर्फ आपने क्यों की, कैबिनेट को क्यों नहीं बताया, अडानी का केस बंद क्यों हुआ, Epstein - ऐसी बहुत सारी चीजें हैं जिसपे खुलकर डिस्कशन होगा।
आ जाइए, डरिए मत।
दूध का दूध, पानी का पानी कीजिए।
आप एलिगेशन लगाते हैं ना मेरे ऊपर, आ जाइए, सब साफ हो जाएगा।
बाकी चीजों पे आपने defamation केस लगा रखा है, शायद 45 defamation केस मेरे ऊपर हैं।
मगर इस डिबेट मामले पे, foreign कनेक्शन पे आप defamation नहीं लगा रहे हो। लगाइए ना, वहाँ भी डिस्कस करेंगे, खुल के। ये सब डिस्कस करते हैं, आइए।
हैप्पी दिवाली, मोदी जी।
It’s been 24 hours since I challenged you, Modi ji.
You say the Opposition is anti-national. Choose any stage, any venue, and debate me on this - Who is anti-national? Who has connections with foreign forces? And what is the nature of those connections?
Let’s do this as soon as possible. Daro mat!
We’ll both be there. The media will be there. The youth of this country will be there. There will be a table. You’ll sit on one side, I’ll sit on the other. You can appoint whoever you want as the judge. You can even make Mohan Bhagwat ji the judge.
There will be questions about the financial system, the financial structure, how the US deal happened, why it happened and why so quickly, why you made the deal alone and didn’t inform the Cabinet, why the Adani case was closed, Epstein - and many other issues. We’ll discuss all of them openly.
Let the truth come out. Let everything be crystal clear.
You make allegations against me, don’t you? Come forward. Everything will become clear.
You’ve filed defamation cases against me over other matters. I think there are around 45 defamation cases against me.
But on this issue of the debate, on the question of foreign connections, you haven’t filed a defamation case. Go ahead, file one. We’ll discuss that openly too.
Happy Diwali, Modi ji.
Paradox of politics: you win when your opponent messes up, not on your own merits
What do regular Americans worry about in the first five minutes after they wake up in the morning? They worry about their kids. Their bills. How they’ll afford to pay the mortgage and fill the tank. They don’t worry about Iran or renovating the White House or any of the numerous other things Trump spent the last year rambling about. Republicans did not speak to the issues that directly and immediately impact regular Americans. They did not present a compelling plan to deal with those issues. That’s why they’re getting crushed. And they deserve it.
Pushkar Ghanekar retweeted
When it comes to hiring, it’s easier to tame a reckless hustler than it is to instill that dawg in someone
RSI = Rapid surgical improvement?
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.…
Pushkar Ghanekar retweeted
Paperclip Datasets is live! Describe the table you need, and agents collect the data from papers, patents, trials and regulatory documents. Each value links to its source passage so you can check it.
Try it: paperclip.gxl.ai/datasets
Read more here: gxl.ai/blog/introducing-pape…
Pushkar Ghanekar retweeted
Can we help discover new, unknown small molecules using modern generative AI and scientific agents?
Excited to share our work, led by superstar @zyxiong123, on ChefNMR2🧑🍳⚛️, a new approach for structure elucidation from NMR spectra!
Paper: chefnmr.cs.princeton.edu/
🧵 1/
Pushkar Ghanekar retweeted
A ‘hello world’ milestone for Project Suncatcher 🚀
After launching our prototype satellite a few weeks ago, we’ve now successfully sent a command up to orbit, booted up the TPU, ran the prompt, and received our first complete response back on earth - a haiku!
Pushkar Ghanekar retweeted
Last night an AI agent performed brain surgery on a mouse.
We gave it 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorize surgery rig with cameras and one goal: reduce seizures.
It picked the drug. It picked the brain region. It derived the coordinates from three papers. Then it drove the stereotaxic arms into the brain and injected 250 nanolitres while watching the entry site through the rig camera for bleeding.
A licensed vet gated every irreversible step and held a physical kill switch to stop the AI agent.
Thread on what it decided, what it saw, and the moment it refused to let me override it.
[Writing this on the spot out of excitement, but I've been 18hrs in front of my laptop coding and coordinating this, so keeping it brief and descriptive, more cohesive thoughts on this soon...]
Pushkar Ghanekar retweeted
This is very needed. Talking to models like Claude about anything that has an overlap with an area like legal, healthcare, etc is mind numbing. They’re so desperate to hedge and steer away from grey area that they often are really avoiding the real question.
Meta's @Muse answers questions very differently from ChatGPT or Anthropic.
It cites more sources but uses far fewer words and tables, and doesn't try to ask a follow-up.
This short-form content is designed for dealing with tasks rather than continuing a conversation, and frankly, it's far less exhausting to read. This speaks to a really well designed AI experience and its so different from the base LLMs.
Great study by the AirOps team 👇
gg @alexandr_wang
Pushkar Ghanekar retweeted
Proud of TEAM Lilly, we are executing well. While 2026 is setting up to be a year to remember I am even more excited about our future…
- 4 Incretin projects in phase 3 now, raising the bar of what’s possible in weight loss, reducing side effect burden and radically expanding indications caused by obesity. We are building the tools to solve overweight and obesity, and hopefully the 200 terrible diseases it causes, at a global scale.
-Rapidly expanding Cancer portfolio in B cell driven disease, Breast and Ovarian cancer as well as Lung and more.
- Treating or preventing the worst neurodegenerative diseases we know
- Unlocking the potential of one and done therapy for common diseases like lymphomas, wet AMD, childhood deafness and even something as common as high LDL cholesterol.
The future of biomedical technology is going to amaze.
Today we announced third-quarter 2026 results. Every quarter is a chance to show what a medicine company can do for people: discover what's next, make it at scale, and get it to the people who need it. go.lilly.com/lkq1zh
Pushkar Ghanekar retweeted
This is Genie 4, our most realistic and interactive world model yet. 🌐
Explore dynamic environments in real time to see how the scene responds, sounds, and how the people in it move, react, and speak.
Pushkar Ghanekar retweeted
We're about to enter a period where not reading or writing becomes as ubiquitous as smoking once was. We can hope people will eventually realize this is bad for them, as they did with smoking. But this is far from certain; the cost may not be as obvious.
Pushkar Ghanekar retweeted
The problem isn’t that AI is intelligent. It isn’t. The problem is that too many adults (but not all) are at once mechanistic, imitative, unimaginative, stupid, formulaic, influenceable, & gullible, arrogant beyond their skill set, & too expensive to maintain.
Pushkar Ghanekar retweeted
Frontier labs are advancing the capabilities in AI for biology but preventing the good guys (scientists, academia, non-profits, health agencies,...) to use, study and understand it with widespread but shallow guardrails that the bad guys are bypassing easily (blocked me to proofread this tweet for example 😅).
This will automatically create massive assymetry and risks in the coming years, just like in cybersecurity, especially now that frontier labs are building wet labs (HF cybersecurity scenario for biology when?).
This is why we created Carbon-X, the best open-source DNA model by midtraining Qwen3-14B on 300B tokens of DNA, English and interleaved data. We also used a hybrid tokenizer that tokenizes English with BPE and DNA with a 6-mer tokenizer.
As a result, Carbon-X outperforms all open language models we evaluated on five benchmarks of DNA understanding and generation. This is a 27-point improvement over Qwen3-14B, showing how much a language model can gain from DNA training. It is also API frontier level at 1,800x less cost than GPT 6.1 Sol!
We hope Carbon-X helps researchers explore DNA in new ways. Releasing it openly means anyone can study its capabilities, limitations, and risks. Let's make the AI biology frontier open-source to make it safe and impactful for the whole world (not just a few companies).
Apache 2.0 Model on @huggingface: huggingface.co/HuggingFaceBi…
