Compute dealer
United States
Joined September 2016
- Tweets8.3K
- Following1.3K
- Followers3.5K
- Likes17.5K
Happens to the best of us
We experienced an outage at Coinbase last night, which is never acceptable. The root cause was a room overheating in an AWS datacenter when multiple chillers failed. We design our services to be redundant to downtime in any one AWS Availability Zone (AZ), and most of our systems worked this way last night, but not all.
Our centralized exchange did not. Exchanges have unique architectures that optimize for latency and co-location of clients. It is possible to make exchanges resistant to AZ failures, but this can introduce latency delays that are not desirable along with breaking customer co-location. Given this incident, we'll revisit these tradeoffs to ensure we're giving you the best possible venue to trade. At a minimum, the duration of an outage should be able to be reduced considerably when an AZ move is needed.
Thank you to the AWS and Coinbase teams for working through the night to mitigate the issue. We’ll share the detailed technical summary once it's ready.
dylan retweeted
Wow.
The White House just announced that grid infrastructure is essential to national defense.
This includes transformers, transmission lines and conductors, substations, and high-voltage circuit breakers.
Companies working to electrify America will have a big tailwind.
dylan retweeted
Half of America's AI data centers planned for 2026 are delayed or cancelled. They're waiting on transformers. I build chemical plants. Transformer prices have tripled in the last four years. Lead times are 2 to 4 years. Each new plant we build competes with AI data centers for the same grid equipment. Every large power transformer in America runs on grain-oriented electrical steel. It's made by rolling iron and silicon together until their crystals align in one direction. No other alloy works at utility scale and only one US company makes it: Cleveland-Cliffs. The average large power transformer on the grid is 38 years old. Service life is 40. Amazon, Google, Meta, and Microsoft committed $650 billion to AI infrastructure this year. Nvidia's most expensive GPU is useless without a transformer.
dylan retweeted
EXCLUSIVE SCOOP: I joined the wonderful Labs team at OpenAI. I am learning a lot and have l enjoyed it very much.
dylan retweeted
getting ready to ship this next batch of openhome dev kits 💪
Major sound driver upgrade. Best speaker you’ve ever heard.
if you build agents or use openclaw, I’ll get you one for free. need feedback. comment or RT what you want to build
dylan retweeted
VCs are max-information-extracting social manipulators wired to digest every conversation to go into their head or firm's CRM
The personal computer has always been a node for self improvement, but mostly for humans. Now it will be for the computer itself.
A decentralized AI computer is an exponential learning machine.
Each node can discover a way to improve itself. Test its theories. And report its results with each other.
The whole network will improve together.
Exponentially quickly.
If you are going to compete with it you have only days to convince same number to load it, even if your system learns faster at first it will never catch up.
@openclaw is all that.
The most disruptive computing system I have ever seen turned on.
@karpathy was right to hype it up.
dylan retweeted
Exciting to see how our cognitive architecture is beginning to transform not only enterprises but also individual creators and builders like @Scobleizer.
We are currently wrapping a funding round with some incredible investors and have just hired our first engineer.
Soon, anyone in the world will be able to work with our agents and experience what real cognitive intelligence can do.
Your Clawdbot can not do this report.
Only @blevlabs can.
Need a job in AI? Need to know what inventions were shared today? What new things were announced?
No one else has a report like this.
Brayden Levangie and I are doing a variety of reports like this to do early explorations for a new kind of media business with his cognitive architecture and my lists, which are the most complete of tech industry here on X.
What would you like us to study about the tech industry here on X?
Personalized news is here, but not evenly distributed. This report costs us about $100 to do.
Not even Grok can do this.
Getting to jam on new technology with @Scobleizer is something I've dreamed about for ~10 years. Feeling blessed and inspired.
Lately, I've been getting invited to extraordinary dinners.
Tonight's was at the AGI House in San Francisco, where I met founder after founder who blew my mind. My mind is mushy right now.
It's clear the world is speeding up. I heard about:
1. A construction robot, run by a woman genius, that ties rebar (the metal in the middle of concrete) six times faster than a human being
2. A new kind of physics engine that blows away any other
3. A guy making Korean food with robotics
4. A guy making a new kind of AI computer.
What a world we're about to get!
dylan retweeted
study graph theory.
because most real problems are graphs pretending to be something else.
• networks.
• dependencies.
• flows.
• constraints.
• social systems.
• supply chains.
• compilers.
• routing.
• biology.
• markets.
• the internet.
• your brain.
• all nodes and edges.
graph theory trains you to stop thinking in sequences and start thinking in structure.
what connects to what. what breaks if this node fails. where the bottleneck actually is.
once you see the world as graphs, you stop solving symptoms and start redesigning systems.
learn trees, DAGs, shortest paths, flows, cuts, matchings.
not for exams. for clarity.
graph theory doesn’t make you faster at coding.
it makes you harder to fool.
Now do it but for datacenters
AI driven Mechanical Design.
I dare say this is now possible.
I can now see how this can be implemented at any engineering company today.
There is an API stack that now exists that connects parametric CAD, parametric multi-physics simulation, intelligent interpretation of results, report writing and agentic decision making.
code.claude.com/docs/en/gith…
github.com/topics/onshape-ap…
github.com/SimScaleGmbH/sims…
I doubt this can design entirely new things in 2026, but I expect this is already the fastest way to explore mechanical design optimisations even of moderately complex systems.
Now, several entire workflow loops are still missing, there is no Design for Manufacture, Quality Control, no context for real world capabilities
But the executive kernel for design and manufacturing is already there. Which is absolutely extraordinary.
The question for a lot of boards in a few years time will be: “do we build a proprietary system, or do we buy turnkey?”
It’s a different answer for different organisations.
But AI + APIs are going to replace the internet and they are going to replace knowledge work.
2016-2025 was a period of incredible uncertainty around what the future looked like, but the fog is quickly lifting and everything is different now.
