Category: Encoded VC

Jake’s post on bandwidth trading and today’s push to financialize GPU compute is one of the better things I’ve read this year. It hits close to home: I was in the room for a small piece of the first act. One of the first board meetings I joined at Battery, in 2001, was for Cinta…

On September 20, twelve days after Meta launched its Muse personal agent, Amazon shut it out of Amazon.com. Muse users trying to shop got a popup saying an unauthorized AI agent was violating Amazon’s Conditions of Use. Amazon’s stated reasons are worth reading closely: Meta never told Amazon Muse would be shopping its store, the…

Bandwidth is cheap. The bytes you have to hold in flight are not. Chris Zeoli’s piece on why AI is a memory / storage workload is a great read. Inference is stateful now, that state has to live somewhere, and the memory-storage boundary is where value is moving. I agree with all of it. But…

When we talk about GPU performance, we obsess over throughput. More tokens per second. Faster inference. But we’re missing a critical insight: the relationship between power and performance is deeply nonlinear, and that nonlinearity has massive implications for data center economics. Provisioned power is the long pole in the datacenter tent. Stop wasting what you…

Sometime in 2017, Reid came to me with a pitch for a company called Packet. I loved the name, but when he said it was a bare-metal hosting business, I replied with the obvious: they compete with AWS. He shot back, “Yes, that’s the point. And the guys are amazing.” I went to the meeting,…