Founder of Sealing Technologies (acquired by Parsons 2023); Prior soldier; Cyber, Systems Engineering, & Business. Currently exploring AI & investment opps.
United States
Joined February 2022
- Tweets2.8K
- Following250
- Followers783
- Likes6.8K
Pinned Tweet
Excited to introduce the platform we've been working on for the past year!
@Patapsco_AI believes you should be able to run your own intelligence. Running AI on a single node is easy. Running it on a cluster is harder. We're giving the community a way to run multiple replicas of very large models across a cluster, with KVCache-aware load balancing, API aggregation, VSCode extension, a custom-built stream caching filesystem, and a bunch of other goodies.
You can run it on a single node DGX Spark, or a 100-node B200 cluster.
We're making it freely available for anyone to download and use at home or at work.
If you already run Kubernetes, you can deploy it directly, there is a tool (pailab) to setup a cluster as a service.
If you have any issues, join our slack (slack.patapsco.ai) and hit us up.
I would consider this a beta-release, but it's been fairly well tested. Would love to know what people think.
Not sure how accurate this is, but I do know "Conviction works both ways".
Ultimately, you are either right or wrong. Having conviction when you are wrong is just stubborn stupidity. But no know advertises that.
Observe, experiment, learn, adjust, repeat. That is the formula. That is what you should have conviction in.
Leopold Aschenbrenner might genuinely be the unluckiest motherfucker alive.
After getting absolutely destroyed in July and being forced to liquidate much of his AI portfolio, he reportedly came back swinging in early September, buying hundreds of millions of dollars worth of options premium.
Look at how those stocks have performed since September 8:
$CRWV: -37.4%
$SKHY: -25.8%
$SNDK: -23.8%
$DRAM: -22.9%
$BE: -18.1%
$INTC: -9.4%
$AMD: +9.1%
Six out of seven are underwater, with an average decline of roughly 18.6%.
And remember, we're talking about someone who reportedly rebuilt exposure through leveraged options, not just common shares.
Imagine surviving a massive forced liquidation, watching your favorite stocks recover, deciding it's finally safe to get back in, and then getting hit by ANOTHER brutal selloff barely a month later.
The worst part about options is that you can be absolutely right about the long-term AI infrastructure thesis and still lose a fortune because your timing was wrong.
I genuinely feel for the guy.
The market has absolutely no mercy.
Prepping for KubeCon & CNCF in Salt Lake City! Can't wait to show off our platform.
Less than 30 days away from #KubeCon + #CloudNativeCon NA and are excited to join our industry colleagues November 9-12 in Salt Lake City, Utah! Stop by Booth 373 to check out how our sovereign AI platform will let your LLM garden grow and enter for a chance to win a DGX Spark!
This is not a bubble pop... this is a slow deflate, which is a good thing. It'll help soften the blow over the next 24-36 months.
As someone who started smoking at 13 and have been addicted to nicotine for 30+ years (still chew the gum everyday)... Cancer is NOT the only reason not to start. It's a prison of the mind and body.
Bought this 18 months ago on Amazon for $320. Forgot about it and just found it on my shelf. The exact same drive it listed for $950 now.
I knew things were rough, but didn't realize they have gotten this bad.
I keep coming back to:
"Everyone should have equal access to intelligence in order for our system to function correctly."
Personally, I don't see the point of the consciousness argument.
It's like arguing whether physics in a computer simulation is "real physics". It doesn't really matter.
These models find and learn correlations in human language (or images, or audio, etc.) and add a probabilistic sampling to enhance search.
On the subject of pain: They don't have a central nervous system, so they don't "feel pain". They understand what the term "pain" is because they correlate it to other concepts. They can convince you, with enough conviction and knowledge, that they know pain, but don't confuse simulated representation with real electrical systems in continual learning organisms. Very different things
With that said, torturing a simulated animal or person, simply for the pleasure of it, is not healthy for to the aggressor's mind and should 100% be avoided, shunned, and shamed. Not because it hurts the model, but because we, as humans, have a hard time telling the difference and therefore are exercising immoral muscles that should not be strengthened.
Remember that your bio-neurons are learning all the time, so giving them the right data is critical.
As the possibility of machine consciousness is doing the rounds today, here is my PhD thesis, published last year, about that very topic.
TLDR: I am reasonably certain Claude isn't conscious.
dspace-prod.anu.edu.au/bitst…
Not sure what was said at that WH summit, but it seems to have worked...
Lots of discussion out there about our next model(!), so I wanted to give an early look as soon as possible. Introducing Gemini 4 Argon!
It shows frontier performance in complex workflows, cyber defense and software engineering. Teams are using it extensively at Google, from coding to quantum computing, great feedback.
Here’s a look at the benchmarks:
Wait till they find out about Kali Linux. They will definitely want that OSS project banned along with this model.
The reality is that it did WAY more for defenders and training blue teams than it did for attackers.
There are more good-guys than bad-guys. So it's net positive. xitter.cf/YGandelsman/status/210…
Anyone else feeling like there are already too many AI protocols and solutions to figure out?... and more are coming out every single day.
They're really trying to find a good use case for the Bluefield cards. Maybe this is it?
NVIDIA Open Agent Safety Platform Reference Design combines NVIDIA OpenShell and NVIDIA Sentry.
OpenShell is an open-source secure runtime that gives AI agents clear, enforceable boundaries. It traces their actions and enforces policy as they work.
NVIDIA Sentry delivers added layer of security with hardware-based enforcement on NVIDIA BlueField, continuously monitoring agent activity through a trusted telemetry and detection pipeline and enabling millisecond-scale containment and quarantine.
IMHO, Google & Meta will be two of the big winners of the AI boom, but not due to their AI, due to marketing and ads.
As software gets easier to build, and new apps and SaaS come online from makers, they will want to market and differentiate their new product.
Self-serve marketing platforms will increase in demand.
Putting competitive products in front of customer's eyes is something the trad platforms already do well and have an entire industry built around them.
Even among people that build and live AI, no one likes to be reached out to by a sales or marketing AI agent.
Sorry, but the "AI will replace sales" ain't gonna happen.
