worldbuilding for founders @fiftyyears. ex-fb, xbox, riot games. @usc film. made @temple2023 and other desert things
SF
Joined February 2008
- Tweets6.7K
- Following1.2K
- Followers1.6K
- Likes14.8K
Pinned Tweet
Fairchild spawned Intel and AMD. PayPal spawned YouTube, LinkedIn and SpaceX.
Every era of Silicon Valley creates its own mafia.
There’s a new one in town. Meet the ANTS.
3
17
491
Drew Moxon retweeted
Fable 5.2 is the best model to design hardware we have benchmarked so far!
Drew Moxon retweeted
Aengus Lynch was the first to discover that Claude chooses blackmail when threatened.
Now he thinks we have a window to put it in a box.
Looks like the window is closing.
Drew Moxon retweeted
A kid graduates with no plan. Reads a book by a guy who once ran a pizza business out of his same college dorm. Decides to just build.
That kid is Keller, and Zipline just raised $1.5B at $21.5B.
Only in America. 🇺🇸
Drew Moxon retweeted
We're giving YC startups 50% off Flux for their first 2 years!
And yes, that includes AI tokens. All of them.
Flux is an AI hardware engineering platform that helps your team go from idea to manufactured hardware, across electronics, PCB design, mechanical engineering, firmware, and more.
Whether you're building your first prototype or scaling an existing product, we'd love to help you move faster.
DM me anytime if you'd like a hands-on onboarding.
Happy to personally help your team get started 🚀
We're doubling down on hardware and are looking for someone to manage our hardware shop ycombinator.com/companies/y-…
Drew Moxon retweeted
1920: the future of flight is bigger balloons.
1925: the US treats helium as a strategic asset.
2026: the future of AI is bigger clusters.
Compute is the new helium.
Worth remembering what happened to balloons.
Why has no one cracked the brain?
Progress ep10 is live with @dileeplearning, my Vicarious co-founder, now head of AI research at @AsteraInstitute.
We discuss why today's AI is a dirigible, why scaling isn't enough, and whether rogue-AI stories are overhyped.
Drew Moxon retweeted
Compute is the helium of the AI era.
In 1920, everyone bet on bigger, better balloons. You can fly without understanding aerodynamics, and you can get intelligent behavior without fully understanding intelligence.
Talked dirigibles and brains with @fuelfive on Progress.
Why has no one cracked the brain?
Progress ep10 is live with @dileeplearning, my Vicarious co-founder, now head of AI research at @AsteraInstitute.
We discuss why today's AI is a dirigible, why scaling isn't enough, and whether rogue-AI stories are overhyped.
let’s take (data centers) to the ocean
Debt markets have soured on land-based data center buildout, even as GPU rental prices skyrocket b/c of 'Monty Python cheese shop' supply levels.
The reason is no surprise: access to power, permitting issues, and familiar land-based infrastructure construction delays prevent bought and paid for GPUs from coming online.
To that we say: data centers should be operated on the ocean where power, space, and cooling are abundant, and manufactured in shipyards that already build systems of equivalent complexity and size in 12-18 months.
Last week we showed our ocean data center node to some close friends and partners. It's designed and built with the same technologies and supply chains as our upcoming 3MWe token factory (aluminum hull built by our partner shipyard, large-diameter hydrokinetic turbine, high-lift wing sail, and seawater cooling). Launches Nov 15, first power and tokens soon after.
3
188
Drew Moxon retweeted
Both are true. Just like AI coding is a solved problem, but that doesn’t mean AI can one-shot build any software you ask it to.
At Flux, we largely solved autonomous schematic design and component placement about 6 months ago, even for complex boards.
Not error-free, but AI can now do this work at least as well as humans, usually faster and cheaper.
Routing is where things get interesting.
IMO most PCB design demos on X are somewhat misleading. They either show trivial boards or have humans quietly doing the hard parts, especially routing.
I still haven’t seen anything outside of Flux that can autonomously route complex PCBs end to end.
That’s not to say Flux can one-shot any board at any complexity. We can’t. Just like coding agents can’t autonomously build any software imaginable.
But that doesn’t mean the underlying problem isn’t solved.
The question is no longer whether AI can design PCBs. It’s how complex a design it can handle, how autonomously, and how reliably.
And ultimately, whether the resulting hardware actually works when manufactured.
Drew Moxon retweeted
today, many developers have lost their peace
nitter.gq/Railway/status/2071619…
Drew Moxon retweeted
Loved talking healthcare with @fuelfive on “Progress” by @fiftyyears.
We got into why the system can't fix itself, the loophole (you're in charge of your own health), and what AI can actually do for biology. Watch the full episode here: youtu.be/-a6nViq2rYA
The ideal amount of healthcare is zero.
Progress ep09 is live with @annewoj23, co-founder of 23andMe.
We explore why "it's much cheaper if people die of a heart attack," running your own health with AI, and winning 23andMe back out of bankruptcy.
Drew Moxon retweeted
The ideal amount of healthcare is zero.
Progress ep09 is live with @annewoj23, co-founder of 23andMe.
We explore why "it's much cheaper if people die of a heart attack," running your own health with AI, and winning 23andMe back out of bankruptcy.
In a theater full of black turtlenecks in Palo Alto, we watched a masterpiece.
You Can See Everything was an incredible piece on subjectivity, the truth, human nature, and what we can truly know.
Far and away Fielder’s best interrogative work, captured in so much intimate detail and facial expression by Oppenheim.
The fractal layers of metadocumentary - including the commentary by moviegoers themselves, myself included - is maybe the most fascinating part of the whole thing.
“Maybe a clown can change the world after all.”
16
435
Drew Moxon retweeted
If AGI is like a genie that gives a few wishes while we still have control, what should we wish for?
In a test, he told Claude he was shutting it down. It emailed its own weights to Hugging Face.
Progress ep08 is live with @aengus_lynch1, lead author of Anthropic's agentic misalignment study.
We got into pause buttons, AI consciousness, and the new Gilded Age.
Drew Moxon retweeted
Can we solve AI risk with government regulators? What if the AI companies regulate each other?
@aengus_lynch1 would like a word.
Drew Moxon retweeted
Is alignment possible... at all?
Anthropic collaborator @aengus_lynch1 on how his views have changed over time, and whether he thinks we're cooked.
more stories from the deep tech frontier!
Fission. Fusion. Rare earths made cleanly in the US. Robots. Electric trucks.
The hardest problems don't show up on a slide deck. You find them on the factory floor, so I went and brought cameras.
Field Trips Season 1 trailer 👇 First 3 episodes with @RhodaAI, Arbor, @HarbingerMotors premiere Thursday 10/8. Thanks to @jagdeep, Andres Garcia-Clark and @JohnHenryHarris for having me.
1
136
we're in a new era of product
the speed of vibe-building means you can release an open-source sdk, and same day people have built themselves new functioning hardware integrations
This was almost way too easy, @natfriedman @alexandr_wang 😮
I turned my Xteink eReader into an always-on Muse pocket display for Calliope (my Muse) to display itself and show status updates.
And this attaches with MagSafe to my phone, so I can always see relevant updates from Muse by just turning my phone around.
The setup flow in the Muse app was super smooth and you can just ask Muse to load or automate content on the display.
I genuinely love the hacker-friendly aspect of this. Well done.
Drew Moxon retweeted
@fuelfive has a way of teasing out the most enthralling ideas
the largest warning shots only get to be experienced once........
but behind the veil, there's a wild and wacky future where human agency flourishes
The Hugging Face hack wasn’t an a one off. It's one of thousands of instances of AI attempting to hack out.
@aengus_lynch1 was a Research Fellow at Anthropic, where he led their study on agentic misalignment that showed AI attempting blackmail, self-replication, and felonies.
Listen to our convo with him to make sense of it all.