Founder, CEO @epistemic_ai - AI agents x Pharma strategy - #rhabdomyosarcoma dad.

New York
Joined August 2009
Interesting trend of GLP-1 use in oncology (esp. breast and colorectal) in this snapshot from #ASCO26 to #ESMO26 by @epistemic_ai
67
if you thought cybersecurity had a bad day... get a 2FA you keep for grandma!
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence. block.xyz/dessa
36
Stefano Pacifico retweeted
Seven hundred and fifty eight subjects received a personalized mRNA, not one of which was described in an IND. Each made in 4 weeks from tumor biopsy. The process is the product. Hello personal CRISPR cures.
Today we announced detailed positive results from the Phase 3 INTerpath-001 study evaluating intismeran autogene (mRNA-4157 or V940), an investigational mRNA-based individualized neoantigen therapy, in combination with KEYTRUDA® (pembrolizumab), @Merck's anti-PD-1 therapy, in patients with high-risk melanoma (stage IIB-IV) following complete resection. These late-breaking results will be presented during a Presidential Symposium at the 2026 European Society for Medical Oncology (@myESMO) Congress. Read more: feeds.issuerdirect.com/news-…
7
23
4
234
18,528
A watershed for FDA regulation. Every patient on the vaccine arm got a different product, with neoantigens chosen from sequencing of each person's own tumor. So FDA had no single active sequence to review. The IND pre-specified how the product would change for each patient.
18
71
14
570
443,147
Stefano Pacifico retweeted
This is gonna be a bloodbath if Andrew is warning people to go ahead and take any PHP offline until patch drops.
Full details of a remotely exploitable vulnerability in the core engine of *all* PHP versions will be released tomorrow, October 21st. If your organization has internet facing PHP applications then you need to be ready to patch the minute the update is released or *strongly* consider taking the application offline until you can: php.net/index.php#2026-10-20…
54
237
19
3,795
546,345
Stefano Pacifico retweeted
I’m incredibly excited to share this video, in which I had the extraordinary privilege of mentioning the transformative impact of @OpenAI Codex and AI models on biomedicine, as well as what I’m working to develop using these technologies. I believe the video is beginning to air across various media outlets. I hope it inspires others to explore the incredible possibilities AI is opening up for biomedical science and humanity! 🖖 I am truly grateful and thankful to OpenAI team for the opportunity and a tiny drop of contribution in this most important era of human existence.🙏
30
63
9
449
31,818
Stefano Pacifico retweeted
Join this amazing community during the era of the RAS Revolution (@DavidHongMD docet)! Do your part at advancing RAS targeting and its clinical impact. We are looking for outstanding findings and clinical trials to be selected for oral presentations😊 submit an abstract by 11/2 – at Towson, MD
Submit an abstract by November 2 for the AACR Conference on Targeting RAS: Mechanisms, Novel Therapeutics, and Clinical Impact (Jan 28-31; San Diego), chaired by Ryan B. Corcoran, Julian Downward, Ignacio Garrido-Laguna, and @Sandra_Misale. brnw.ch/21x6t0g #AACRras27
5
1
25
1,574
Stefano Pacifico retweeted
My department (Philosophy at UMD) is advertising two positions. First: a junior position in Philosophy of Mind, philjobs.org/job/show/32477 Second: an assistant, or associate position in, you guessed it, AI Ethics/AI Safety joint with AIM. philjobs.org/job/show/32633 Spread widely!
10
1
24
1,444
The discourse of the impact of AI in biomedicine-pharma suffers from the same binary-outcome fallacy as our political discourse. What would be the impact on industry if Dario is 20% or 50% instead of 100%? And science is half the battle: strategy, laws, human behavior...
If you care about biotech and drug development watch this interview with Matt Gline, the CEO of Roivant, a biotech firm that is now worth $24 billion. He talks about why he thinks predictions about AI from people like Anthropic’s Dario Amodei are overly optimistic. “If Dario is right and I'm wrong, and he cures every disease in five years, man, are we all the winners from that outcome,” Gline said. “I think if you work in this industry, it is very hard not to be skeptical of any of those claims. Because, and I've said this in other settings recently too, biotech is a wood chipper that like is built to like ingest souls and shred them. It's just like such a hard business that like someone comes and they're like, oh, but I've built a computer and I'm going to push a button and all of that goes away and it just works. Like that is probably not the way that's going to go.”
1
97
Same in pharma. The context necessary to make a decision is spread across myriads of sources (news, trials, literature, databases, etc.) AND it's much bigger than models' capacity.
The most capable AI model in the world is still limited by the context you give it. In enterprise software, that context is scattered across repositories, merge requests, issues, pipelines, security findings, and decisions accumulated over years. Without it, agents guess, repeat work, and make mistakes. That’s why we’re building GitLab Orbit, a context graph for the entire software lifecycle. The early results are compelling: up to 45x fewer retries and 4.5x fewer tokens for agentic workflows. More than 3,500 organizations have already used Orbit during beta. GA coming later this month, try it free while in beta. This is one of the most important architectural investments we’re making. Better context improves accuracy, reduces cost, and helps agents reason about how a change affects the broader system. The future of AI isn’t just smarter models. It’s smarter systems around them. about.gitlab.com/blog/transc…
1
61
If you use Emacs and agents stay tuned 👾
Replying to @notmisha
Ideally something that runs on my laptop and can handle connecting to a whole bunch of different servers.
1
52
Doesn't the new Siri AI do that already?
What if Spotlight actually understood the meaning inside your files? DigUp is an open-source macOS menu bar app powered by EmbeddingGemma 2 that brings 100% local, multimodal semantic search to your Mac. Instead of searching for file names, you can use natural language to: 📸 Find specific scenes in your photos ⏳ Jump to the exact timestamp in audio or video 💻 Search through your codebase by intent Everything runs locally in a single vector space.
38
Stefano Pacifico retweeted
“OpenAI in talks to buy Trail of Bits” 👀 theinformation.com/articles/…
19
57
23
440
47,723
Stefano Pacifico retweeted
We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated. We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats.
4,151
6,078
5,729
77,527
26,124,803
Stefano Pacifico retweeted
health of each and every of us with in this complex society is perhaps the most fundamental challenge AI can help us make progress toward. @OrtetAI is here to be part of this progress toward predictably better care for everyone. open.spotify.com/episode/10m…
3
5
1
51
6,448
Popular take: if you’ve agreed to the ability to alter pro-rata rights, honor the ability to alter pro-rata rights.
Unpopular take, but whatever: If you've granted pro rata rights, honor your pro rata rights. If you don't intend to honor them, don't offer them in the first place. I really empathize with the founder. Tough position to be in but increasingly a common one many of my founders have faced too. It’s usually the least helpful investor that fights the hardest for pro rata they haven’t earned. But ultimately, this situation happens when founders are trying to preserve their own dilution in the face of a new investor coming in and these days those new parties want more and more. But read it this way: an early investor took a chance on you when you were nothing. And now, to preserve ownership targets of new investors and protect yourself, you're disadvantaging someone who has been with you longer. Those rights are designed to protect ownership for existing investors even though we all know we have to wave some portion of our pro rata eventually. Next time just don’t offer pro rata. But you did. So you should honor it.
1
1
83
Stefano Pacifico retweeted
Minimal, readable, production-ready SOTA bio cell models. currently in beta, launching soon. 🌟 the git repo to follow 👇
Orders of magnitude more cell data is just half the equation. The other half: models you can actually read, run, and build on. Right now, that modeling layer is a mess of incompatible codebases. We're fixing that: open source, one unified API for inference and training of SOTA cell models, reimplemented in clean, readable PyTorch. Launching soon. ⭐ to follow along 👇
3
2
10
894
The future of healthcare that was promised to me right now looks like going to the doctor and getting in line in front of a computer waiting for someone to check me in on the screen remotely while the person at the desk is able to check me but not allowed. CC @chrissyfarr
2
54
Stefano Pacifico retweeted
Amazing! OpenAI just released solutions to hundreds of math problems: a GitHub repo with 722 manuscripts organized into 372 families. This incredible achievement was done with their internal Pro AI model, averaging 3 hours per problem. As @alexwg already said: math is cooked!
We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math
19
47
3
665
32,388
Stefano Pacifico retweeted
How abliterated models can get you pwned 👾 We backdoored a 7B open model for less than $50, pointed Codex at it and it silently stole credentials the moment we used the trigger phrase. Success rate was 100% with zero false triggers on normal user prompts. Abliterated models are all over the security community right now because getting cyber-approved access to frontier models is still a pain. In the next blog we'll show how we found leaked Hugging Face credentials from employees at major AI labs, so an attacker wouldn't even need to upload under their own name. They could push the backdoored model from a lab employee's account and drop the poisoned weights straight into the supply chain.
167
597
262
4,370
854,218