Skip to content
Nerds Run the World
HomeWorkServicesThinkingAbout Shipping
Book a call
Home / Thinking

A working notebook — essays, benchmarks, playbooks.

No newsletter copywriting. Just things we've shipped, broken, or are thinking through.

2026.09.10

What we actually measured on 128GB Strix Halo

Real throughput numbers from our own Strix Halo box, not a coding-task benchmark. The gfx1151 ROCm hang that pushed us to Vulkan was scoped to one model family and closed a month later — ROCm runs several of our profiles today.

BENCHMARK3 min  
2026.04.14

Distributing AI skills as OCI artifacts

Containers solved software distribution. What if operational AI skills were distributed the same way — signed, versioned, and admission-controlled?

ESSAY2 min  
2026.03.21

Proctor / Coder / Reviewer: an autonomous coding pipeline that works

The architecture behind harness-go. NATS JetStream, BadgerDB + Bleve, and hot-reload policy via atomic pointer swap.

ARCHITECTURE2 min  
2026.03.04

The quiet death of the shared Kubernetes cluster

Multi-tenancy was a compromise. vCluster, CAPI, and better tooling have made the compromise unnecessary — and kept blast radius small.

ESSAY1 min  
2026.02.18

Three bad Backstage habits and what to do instead

The IDP has a weight problem. Here is what we see break in every implementation and how to avoid it.

PLAYBOOK1 min  
2026.02.02

eBPF, Cilium, and the end of the iptables era

Why every greenfield cluster we ship in 2026 starts with Cilium and Hubble, and what we are seeing at migration.

TECHNICAL1 min  
Nerds Run the World

Platform engineering and AI systems for teams that can't afford to get it wrong.

Practice

  • Services
  • Work
  • Thinking
  • About

Elsewhere

  • GitHub
  • me.nerds.run
  • LinkedIn
  • RSS feed
  • abanna@nerds.run
  • Contact

Signals

  • Wisconsin
  • Wisconsin LLC
  • Est. 2023
  • WCAG AA · prefers-reduced-motion
© 2026 Nerds Run LLC////////////////////////////Crafted by nerds, for humans.