Writing

The engineering record.

Case studies, series, and technical notes about how systems were built, why the architecture took its final shape, and where the evidence stops.

Featured
W-04 22 June 2026

Hardware / Case study

Pushing the Limits: Overclocking SPECULAR-CORE

A full system optimisation pass on SPECULAR-CORE: BIOS-level memory and Infinity Fabric synchronisation, a hand-built GPU voltage/frequency curve, cooling tuned around real-time audio noise constraints, and thread scheduling aligned to the 9950X3D dual-CCD layout.

Read featured case study →
Browse
W-08 NEXT CHAPTER // AUGUST 2026
DevOps Docker WSL2 Infrastructure Linux
~8 min read

SPECULAR-CORE: Architectural Recovery

Infrastructure // Coming Soon

This June 2026 recovery record follows a compound Docker Desktop and Proton VPN failure across the local AI stack, then documents the migration from Windows-managed containers to a native Linux Docker engine inside WSL2. The result removed the Windows Hyper-V path from this specific topology without claiming immunity from every future host, bridge, kernel, or VPN failure.

Original Architecture The Failure Chain New Architecture
The next chapter
W-07 27 July 2026
DevOps Cloudflare Workers CI/CD Observability GitHub Actions
~10 min read

Atlas Lab

Atlas Lab — Live Observability // Case Study

This dated June 2026 build log records how pipeline events from eight rollout repositories were added to the Lab ring buffer alongside a pipeline status grid and an API surface map. The current estate has since added named signal-class routing in atlas-notify and moved authoritative public Worker discovery to the fail-closed atlas-api-index registry.

Background The Persist-Only Decision Wiring It Into Every Pipeline Closing the Surface
Read case study
W-06 25 July 2026
DevOps GitHub Actions Cloudflare CI/CD Infrastructure
~12 min read

Atlas Systems CI/CD

Atlas Systems CI/CD — Reusable Pipelines & Hard Gates // Case Study

Twelve consumer repositories were brought onto a consistent, gated, observable CI/CD rollout, with two reusable GitHub Actions workflows defined in atlas-infra and short caller files in each repository. The current estate pins those callers to immutable atlas-infra revisions, so later workflow changes move through explicit, reviewable pin-update pull requests rather than floating into consumers automatically.

Background Reusable Workflows as the Architecture Gating atlas-notify Through workflow_run Hard Gate or Soft Gate The Token Model
Read case study
W-05 23 July 2026
DevOps Infrastructure CI/CD Cloudflare Workers Observability
~12 min read

Atlas Systems Infrastructure

Atlas Systems Infrastructure — Infrastructure Dashboard & Observability // Case Study

Building a live infrastructure dashboard and automated deployment pipeline across ten Cloudflare Workers repos, starting from manual deploys with no audit trail and ending with a self-reporting system that documents its own operation in Discord. A single authenticated event bus accepts three webhook auth dialects, observability Workers proxy GitHub and Cloudflare analytics behind cached endpoints, and identical GitHub Actions workflows deploy every Worker on push, all on free-tier infrastructure at zero cost.

Architecture The Event Bus Observability Workers CI/CD Automation Personal Tooling
Read case study
W-04 22 June 2026
Hardware Overclocking Performance AM5
~13 min read

Pushing the Limits: Overclocking SPECULAR-CORE

SPECULAR-CORE — Hardware Tuning // Case Study

A full system optimisation pass on SPECULAR-CORE: BIOS-level memory and Infinity Fabric synchronisation, a GPU voltage/frequency curve built by hand after a beta tool's batch editor failed, AIO and case cooling tuned around real-time audio noise constraints, and Process Lasso thread scheduling that finally gives Windows architectural awareness of the 9950X3D's dual-CCD layout. Validated through a full overnight stress test with zero crashes, zero thermal events.

Memory & Fabric Sync GPU V/F Curve AIO Cooling Process Lasso CCD Scheduling Stability Validation
Read case study
W-03 27 May 2026
AI / LLMs RAG DevOps Docker
~12 min read

Building a Private AI Infrastructure From Scratch

Ramone — Local AI System // Case Study

How I designed and built a fully self-hosted AI system running five LLMs locally on consumer hardware — zero cloud dependency, zero data egress. This covers the full stack: WSL2 as the hypervisor layer, Ollama for model serving, Docker for containerisation, and Open WebUI as the interface. Includes the boot automation system, RAG pipeline architecture, and the hardware portability decisions that make the whole thing rebuildable in under 30 minutes.

System Architecture Model Selection RAG Pipeline Design Boot Automation Hardware Decisions Workbot Engineering
Read case study
W-02 30 April 2026
Audio Systems Game Dev Unreal Engine 5
~15 min read

Engineering a 15-Stem Dynamic Audio Engine for Competitive Game

SlamPunk — Dynamic Mix Engine // Case Study

A deep technical breakdown of the audio architecture behind SlamPunk — how a 15-stem interactive music system was designed to mirror player performance in real time, the MetaSound graph structure, the Blueprint logic driving intensity transitions, and the spectral carving decisions that keep competitive gameplay cues surgically clear at peak musical intensity.

Stem Architecture MetaSound Design Blueprint Logic Submix Hierarchy Sidechain Ducking Export Pipeline
Read case study
W-01 28 March 2026
Generative Audio Max/MSP DSP Jitter
~14 min read

Building an Autonomous Generative Audio-Visual System

SONIN — Dual-Engine Generative System // Case Study

A full technical breakdown of SONIN — an autonomous digital instrument built in Max/MSP that composes its own evolving music and visuals in real time by continuously listening and reacting to the sounds it creates. Covers the dual-engine architecture, the three engineering problems that nearly broke it, and the DSP solutions that made it stable.

Melodic Engine Granular Synthesis Feedback Loop GPU Visual Pipeline DSP Problem Solving Max/MSP Architecture
Read case study