Important modeling work

1 year ago: "We developed and released tools for automated vulnerability detection. Look how we're using it to secure medical devices!" <everyone shrugs>

Today: "Anthropic's new model automates vulnerability detection, but they are not releasing it because it's too dangerous." <everyone freaks out>

Max Hodak — Restoring Sight, Growing Neurons on Silicon, and Expanding Human Intelligence YouTube video by Juan Benet Podcast

Excited to launch a new podcast on the future of neurotech, computing, intelligence, and more.

First guest: @maxhodak.bsky.social founder & CEO of @science.xyz building PRIMA, a retinal prosthetic restoring meaningful vision for patients with AMD.

www.youtube.com/watch?v=24CM...

What is holding #neurotech back from being one of the greatest breakthroughs of our time?

Infrastructure.

Getting from approval to access takes years — delaying patient access & company capital.

PL Neuro aims to break these bottlenecks. Here's our Roadmap: zenodo.org/records/1924...

What about using AI to control a fusion plasma? 😉 In all these domains, there are plenty of engineering sub-problems that are fertile ground for cross-disciplinary collaboration. But I do agree that saying "we'll use X to solve ALL of field Y" is too vainglorious to have any credibility.

Homepage — PL Neuro

PL Neuro is officially live! plneuro.xyz

We exist to break bottlenecks, accelerate progress in neurotech & NeuroAI, and to invest in innovations that benefit humanity.

PL Neuro will focus on 3 core areas:
- Neural augmentation
- Biologically-inspired intelligence
- Whole brain emulation

1/ Neurotechnology is one of the defining opportunities of this century. We’re already seeing the potential to restore function, treat disease, and eventually expand human capability.

Preprint out arguing that we should build the techology to translate (compile) molecularly annotated connectomes into dynamics. I think this is incredibly important. arxiv.org/abs/2603.25713

Meet the Man Making Music With His Brain Implant Galen Buckwalter says brain-computer interfaces will have to be enjoyable to use if the technology is going to be successful.

Galen Buckwalter, a quadriplegic, received a brain implant in 2024 that is allowing him to create music with his thoughts.

"It is so empowering to be able to do entirely unique things," he says.

I spoke with Galen about the process and the importance of creativity: www.wired.com/story/meet-t...

Meet the Head of PL Neuro, @seanescola.bsky.social.

In his new paper, he introduces the idea of “cognitive dark matter."

A thoughtful framing of one of the biggest open questions in neuroscience. 👇

Governor of Texas just ordered a statewide medical cybersecurity review based on my team's work for @arpa-h.bsky.social Fantastic real-world impact for patients! 👇 gov.texas.gov/uploads/file...

"Russian flu," the pandemic that hit in 1977, bears an evolutionary signature of having emerged from a lab, perhaps as part of a failed vaccine effort. Covid, mpox, Ebola, and other influenza pandemics don't. Here's my story on a new way to trace the origins of pandemics. Gift link: nyti.ms/46N0W33

Monitoring gene expression in the brain with a simple blood test?

In collaboration with @jerzyszablowski.bsky.social & his lab we figured out how to do it nonhuman primates using noninvasive neuroengineering of synthetic serum markers.

nam11.safelinks.protection.outlook.com?url=https%3A...

Connectomics is Method of the Year 2025!!
@natmethods.nature.com
just released the news nature.com/articles/s41... In my perspective piece, I afford some predictions into the future and compare how we are doing vs. genomics www.nature.com/articles/s41...

1 year later and wow, exciting to see big things in the works!!!

Our new paper showcasing molecular connectomics with pan-expansion microscopy is out in @natbiotech.nature.com! www.nature.com/articles/s41...
This wonderful collaboration with @bewersdorflab.bsky.social was led by Ons M'Saad (now founder/CEO of Panluminate) and @allisonphysics.bsky.social. (1/5)

We still need a massive improvement in compute efficiency of binder design tools to make this a financially compelling alternative to CROs for the average user.

Furthermore, the above cost of binder design in the cloud doesn't include the cost of experimental validation of your candidates, which would be needed to achieve the same results as a CRO. Might be another few K in labor and materials, plus time needed to make the thing.

Lots of excitement around open source computational binder design (BoltzGen FTW!), but as far as I can tell it's still ~2x more expensive per target than a traditional campaign with a CRO.

(60k runs/target) x (2 mins/run) x ($5/hr for an A100) = $10k/target

vs ~$5k with a CRO

Incorporating open connectomics data into teaching neuroscience Learn to analyze open neuroscience data and introduce dry lab modules into your existing classes at the Incorporating Open Connectomics data into...

I will be co-teaching a summer course at Allen Institute on connectomics education please apply. Travel support, new connectomics data sets and learning directly from the scientists who built these datasets. Details here: alleninstitute.org/events/incor...

Jaxley: differentiable simulation enables large-scale training of detailed biophysical models of neural dynamics - Nature Methods Jaxley is a versatile platform for biophysical modeling in neuroscience. It allows efficiently simulating large-scale biophysical models on CPUs, GPUs and TPUs. Model parameters can be optimized with ...

I am super happy to share that our project on training biophysical models with Jaxley is now published in Nature Methods: www.nature.com/articles/s41...

It's increasingly tempting to believe frontier models truly understand concepts and their relationships like humans do. Then I get a figure like this from today's leading model and the illusion is ruined. <sigh>

Maybe it's specific to AI tools? As a comp, this non-AI announcement + blog post from last year did much larger numbers on Bluesky, well before the paper was released. People were happy to like and share: bsky.app/profile/andr...

Is there an academic/industry divide in attitudes about using AI to support discovery? I noticed this post has 3.6k likes on X but only 6 likes on Bluesky. It deserves more attention here!

For a decade, my white whale as a neuroscientist has been working to convince public+private funders to "go big" on post-mortem human tissue analysis, so we can define circuit disruptions in disease and just maybe link them with molecular drivers. This remains as tough a sell today as 10 years ago.

Haven't seen Carlos in forever, but reading this gleeful post in his voice made my morning. Super cool work

Today, we're announcing Kosmos, our newest AI Scientist, available today. Kosmos makes fully autonomous scientific discoveries at scale by analyzing datasets and literature, and is the most powerful agent for science so far. Beta users estimate that Kosmos does 6 months of work in a single day.

‘A true genocide’: RSF kills ‘at least 1,500 people’ in Sudan’s el-Fasher Saudi Arabia, Egypt, Turkiye, Jordan and Qatar express alarm as more evidence emerges of mass killings by paramilitary forces in Sudan&#039;s in el-Fasher.

As this widely reported genocide plays out in real time and Arab nations express growing horror, it's sad to see the continued apathy toward it in the West. Silence on both left and right wing social media. No demonstrations, no boycotts... www.aljazeera.com/news/2025/10...

A global mission to restore vision | PRIMA by Science YouTube video by Science

Very excited that the clinical trial results for our PRIMA retinal prosthesis are published today in the New England Journal of Medicine! This is the first time that patients who are blind due to photoreceptor loss have been able to intuitively see again.

www.youtube.com/watch?v=J_qT...