Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet's history. https://t.co/2zX5bHdhsa
@AndrewDamitio I’ve lived in 2 places w UGBs (Portland and Marin County CA) and discontinuity analysis (vs Clark County, Alameda County) show that housing is hardly cheaper without a UGB. Anyway… fare higher? Nah. Need the “mass” in mass transit.
@AndrewDamitio Love the diversity of your ideas and opinions! Seriously! That said, density is great and 2-5 acre lots are great too (rural folks are not the enemy). But the 0.5-acre suburbia… that is problematic on lots of levels so I love UGBs because they break the 0.5 acre McMansion model.
@RobertMSterling@friendlybooks Ok you’re definitely a badass and I trust your gut! Q: would this service work for my large Portland OR HOA? (I.e., a nonprofit corp … also: pls don’t make fun of us for being in Portland — we’re gonna kick the lunatics out sooner or later and then we’ll be the best of America!)
Lost in the debates over AI/data center energy use is that computing becomes less energy intensive over time.
It's called Koomey's Law.
In 1999, Forbes predicted that 50% of 2009 energy use would power computers.
Obviously that didn't come to pass.
https://t.co/KEeQIkVCx7
These days it seems like all the huge data + tech stories are about AI. But Amazon’s S3 Files announcement is pretty darned big. Here’s a very sober take that still says… it’s the real deal.
https://t.co/3pStP4esUf
We have a technology so advanced that 10 facilities can meet all the energy needs of California.
California’s policy is to shut down the one facility we have.
@ericfruits Already done! Now … how can we get the message out to normies who aren’t on X or Substack or Reddit that everyone is gonna pay (renters, owners, all customers of every business) and that (based on current practices) they will see zero road improvement for their money?
The Ray Serve team is announcing major performance improvements for online inference with Ray, increasing peak throughput by 11.1x and reducing latency by 88%.
The key enablers: HAProxy for HTTP ingress and direct gRPC for inter-replica communication. Here's how it all works (+ bonus optimizations) 🧵
In the next version of Ray, you'll be able to see massive performance improvements for throughput and latency in Ray Serve, which is especially important for @vllm_project workloads.
@seiji_________ and folks did a great job helping drive these performance improvements!
@AndrewDamitio Have ya seen the clowncil? Banning humane foie gras (lest it be confused with anything else) then shrimp; requiring single person all gender bathrooms to change their signage to be more welcoming; publicly mocking seniors citizens (multiple times). Too busy for cherry blossoms!