Iihopkinsinihopkins.hashnode.dev·Sep 18 · 4 min readA Type System for GPU Auctions: Turning AI Workloads Into Comparable Cluster BidsTake a cluster request that reads: “64 H100s for training.” It feels precise because the model and quantity are explicit, and the word training supplies a workload. Procurement can send it and provide00
Iihopkinsinihopkins.hashnode.dev·Aug 11 · 10 min readAverage Cloud Costs Break Down Around AI AgentsAverage Cloud Costs Miss the Way Agents Actually Run Cloud planning has always leaned on averages. Requests per second. Compute utilization. Storage growth. Cost per user. None of these numbers is per00
Iihopkinsinihopkins.hashnode.dev·Apr 20 · 3 min readZero Nonsense Guide on Deploying on a Decentralized Virtual ServerI’ve spent enough time on cloud platforms to be suspicious of anything that sounds too clean, too cheap, or too eager to “change everything.” So when I first looked at Fluence Virtual Servers, I wasn’00
