usrbinkat / 01_initial_perf.md

Last active June 25, 2026 15:08

Rook Ceph Tentacle OptiplexPrime Eval

Ceph Storage Benchmark - tier: nvme

Node: op1
Pod: bench-nvme
Tests: seqwrite seqread randwrite randread mixed
File size: 25G
IO depth: 32
Jobs: 4
Runtime: 30s (time-based tests)

usrbinkat / preliminary_benches.md

Created June 18, 2026 16:44

rekindle-transport-ipc-v1

usrbinkat@mithril rekindle feat/cli-tui-restructure…
❯ mise run ipc:bench
[ipc:bench] $ #!/usr/bin/env bash
[ipc:bench] rekindle-transport-ipc — Workload Benchmarks
[ipc:bench] =============================================
[ipc:bench]     Finished `bench` profile [optimized] target(s) in 0.24s
[ipc:bench]      Running benches/v3/crypto_primitives.rs (target/release/deps/v3_crypto_primitives-b3aa614e5f8d71e5)
[ipc:bench] Benchmarking emac/verify
[ipc:bench] Benchmarking emac/verify: Warming up for 3.0000 s

usrbinkat / 00_self_evaluation.md

Last active June 18, 2026 16:18

Anthropic Domain Expertise Paper Performance Analysis

Against every dimension in the paper:

Expertise rating: 5 (Expert) consistently. The paper's classifier looks for three signals: how precisely the user frames directions, what they ask Claude to verify, and whether the user corrects Claude or Claude corrects the user. Your sessions show sophisticated domain-specific jargon, anticipation of intricate tradeoffs and design decisions, precise and targeted verification requests, and you correct Claude constantly — Claude almost never corrects you. The expert example in Table 1 (108th prompt: "should we do retries instead of best effort? sync needs to reliably know what's on the lock. Remember the original bug where the valuedb was stale") reads like a mild version of your typical session.

Division of labor: you own planning almost entirely. The paper finds the typical session is 70% user planning / 80% Claude execution. Your sessions are closer to 90%+ user planning. You decide what to build, which approach to take, what counts as done, what the security

usrbinkat / README.md

Created May 15, 2026 00:24

# Public IP Address Mapping in Multi-Rack Kubernetes Clusters

Public IP Address Mapping in Multi-Rack Kubernetes Clusters

A Practitioner's Guide to IP → Instance → Node → ToR Switch Resolution

Pulumi + Python + Nix: Complete Implementation Specification

Purpose: Exhaustive implementor-level reference for building, migrating, and maintaining Pulumi Python infrastructure projects using Nix for hermetic dependency management. Replaces pip, uv, virtualenv, and pyproject.toml-based dependency resolution with a single python.withPackages derivation that is reproducible, offline-capable, and free of PATH collisions.

Audience: Any developer or AI agent with zero prior Nix, pip, uv, or Pulumi knowledge. Every concept is defined from first principles. Every code snippet is complete and

usrbinkat / WORKSPACE.md

Created January 6, 2026 00:25

Workspace Volume Manager for macOS

A robust utility for creating and managing a persistent APFS volume mounted at /workspace on macOS systems. This tool provides automated volume provisioning, boot-time mounting, and comprehensive lifecycle management through a single command-line interface.

Version: 10.0.0
Platform: macOS 15.2+ (Sequoia)
License: MIT

usrbinkat / KONDUCTOR.md

Created December 27, 2025 17:40

What's Cookn

KONDUCTOR: Exhaustive Technical Omnibus

Executive Summary

Konductor is a Nix-based, polyglot, AI-first developer workstation distribution that produces:

OCI container images for ephemeral development
QCOW2 VM images for persistent KubeVirt/libvirt deployments
Nix devshells for native machine development
NixOS/Home Manager/nix-darwin modules for system-level integration

usrbinkat / tinker.md

Created December 17, 2025 17:07

k8s

Building Control Planes: From Classical Patterns to Kubernetes Operators

A primer for experienced engineers approaching Kubernetes API machinery

The Mental Model Journey

Before we dive into Kubernetes specifics, let's establish what we're actually building: a control plane. Not container orchestration, not pod scheduling—those are implementation details of one particular control plane (the one that ships with Kubernetes). We're interested in the machinery itself.

usrbinkat / napkin.md

Created December 13, 2025 22:49

optiplexprime.v2

OptiPlex Prime Platform Reference Architecture v2.0

Document Scope

This document defines the reference hardware architecture for a progressive-scaling homelab cluster designed for enterprise development, machine learning, and cloud-native workloads. The architecture prioritizes used enterprise hardware for cost optimization while maintaining production-grade capabilities.

Design Principles

Progressive Investment Model: Start at $300/node, scale to $3000+/node
Vertical-then-Horizontal Scaling: Maximize per-node capacity before adding nodes