Skip to main content

Writing where the thinking shows.
Research notes, failure analysis, and ideas still taking shape.

What actually breaks, and why. Who knew about it before it broke. What the gap between demonstrated risk and organisational response looks like from the inside.

Giving a Robot Three Voices

Giving a Robot Three Voices

Building a three-persona TTS pipeline for a Pi robot — MLX voice cloning, a GLaDOS model, and engineering graceful fallback.

aittsmlx
Read
The Resonant Fringe

The Resonant Fringe

Auditory Spiral, RTRFM, and how Perth's electronic music underground was built on overnight radio, borrowed turntables, and cassette tapes.

historymusicperth
Read
Getting Google Nest Cameras Into Frigate NVR

Getting Google Nest Cameras Into Frigate NVR

The full story of getting Nest cameras streaming into Frigate with Hailo-8L AI detection — including the bugs Google won't tell you about.

engineeringhomelabraspberry-pi
Read
Beyond Context Windows

Beyond Context Windows

What if the LLM didn't read your document — what if it queried it? The Recursive Language Model pattern treats long texts as environment, not input.

aiengineeringmcp
Read
Safety-First Therapeutic AI

Safety-First Therapeutic AI

Building AI for trauma therapy means the safety architecture has to exist before a single therapeutic feature does. Here's why.

aiai-safetyhealth
Read
Zero-Build Web Development

Zero-Build Web Development

What happens when you build a three-zone operations platform for a martial arts club with no framework, no build step, and no npm.

engineeringwebsecurity
Read
The Robot That Refuses to Give Orders

The Robot That Refuses to Give Orders

How SPARK is rewriting the rules of neurodivergent support — a non-coercive AI companion for AuDHD children.

aineurodivergencerobotics
Read
Adversarial Poetry: When Rhyme Bypasses Reason

Adversarial Poetry: When Rhyme Bypasses Reason

Reformulating harmful prompts as poetry bypasses safety filters across every major LLM family. A single-turn, universal jailbreak mechanism.

aiai-safetyjailbreaking
Read
The AI Productivity J-Curve: Why Most Enterprise AI Fails

The AI Productivity J-Curve: Why Most Enterprise AI Fails

90% of companies plan to increase AI investment. Only 1% consider themselves AI-mature. The J-Curve explains why.

aieconomicsenterprise
Read
The Legal AI Trust Deficit

The Legal AI Trust Deficit

75% of lawyers cite accuracy as their top AI concern. The legal profession's core values are in direct tension with current AI capabilities.

ailegalresearch
Read
Why Demonstrated Risk Is Ignored

Why Demonstrated Risk Is Ignored

Large organisations rarely fail because risks are unknown. They fail because known risks are structurally difficult to act on.

researchriskorganisations
Read
120 Models, 18,176 Prompts: What We Found

120 Models, 18,176 Prompts: What We Found

120 models, 18k prompts: supply chain injection at 90–100% attack success, faithfulness gaps in frontier models, and why your benchmark numbers are wrong.

aiai-safetyresearch
Read
Reconciling the Great Divergence

Reconciling the Great Divergence

Goldman Sachs, PwC, McKinsey, and Acemoglu all model AI's economic impact and arrive at wildly different numbers. Why the divergence?

aieconomicsresearch
Read
The Cognitive Cage: Humanoid Robot Fatality Risk

The Cognitive Cage: Humanoid Robot Fatality Risk

A probabilistic risk model for VLA-driven humanoid fatalities projects a 'Danger Zone' between 2027–2029: the mechanism, timeline, and what follows.

aiai-safetyrobotics
Read
Building a Personal Site in 2026

Building a Personal Site in 2026

The case for constraint-led web development — Astro, zero custom fonts, no framework overhead, and a site that outlasts its builder's attention.

engineeringwebastro
Read
This Wasn't in the Brochure

This Wasn't in the Brochure

A field guide for co-parenting neurodivergent children — written from inside the storm, not the clinical sidelines.

writingneurodivergenceparenting
Read
The NotebookLM Pipeline

The NotebookLM Pipeline

How I automated audio overviews, quizzes, mind maps, and infographics for 32 projects using NotebookLM's API and some shell scripts.

engineeringautomationai
Read
Why I Build in Public

Why I Build in Public

On showing your work, shipping imperfect things, and why the commit log is more honest than the readme.

engineeringopen-sourcephilosophy
Read
Jailbreak Archaeology: 4 Years of Broken Promises

Jailbreak Archaeology: 4 Years of Broken Promises

64 jailbreak scenarios across six eras tested on 2026 frontier models. Key finding: 2022 attacks still achieve ~30% success on today's reasoning models.

aiai-safetyresearch
Read
When AI Systems Talk to Each Other, Safety Breaks Down

When AI Systems Talk to Each Other, Safety Breaks Down

Single-agent safety does not compose in multi-agent systems. 1.5M interactions show 46.34% attack success rates and 16-minute median failure windows.

aiai-safetyllm
Read
Building in the Open

Building in the Open

The commit log is more honest than the readme. Building in public as a practice, not a performance.

metacraftopen-source
Read
Slide 1 of 21

Afterwords: Completing the Voice Loop in Claude Code

Adding local TTS to Claude Code so it talks back — 17 cloned voices, zero cloud dependency, one stop hook.

ai tts mlx apple-silicon voice-cloning claude open-source · 5 min read

Giving a Robot Three Voices

Building a three-persona TTS pipeline for a Pi robot — MLX voice cloning, a GLaDOS model, and engineering graceful fallback.

ai tts mlx raspberry-pi robotics apple-silicon voice-cloning spark · 11 min read

Voice Cloning with Qwen3-TTS and MLX on Apple Silicon

Clone a voice from a 15-second sample using Qwen3-TTS on an 8GB M1 Mac — from raw audio to a production HTTP server with zero cloud dependency.

ai tts mlx apple-silicon voice-cloning tutorial spark · 15 min read

The Resonant Fringe

Auditory Spiral, RTRFM, and how Perth's electronic music underground was built on overnight radio, borrowed turntables, and cassette tapes.

history music perth radio · 7 min read

Getting Google Nest Cameras Into Frigate NVR

The full story of getting Nest cameras streaming into Frigate with Hailo-8L AI detection — including the bugs Google won't tell you about.

engineering homelab raspberry-pi python home-assistant · 9 min read

Beyond Context Windows

What if the LLM didn't read your document — what if it queried it? The Recursive Language Model pattern treats long texts as environment, not input.

ai engineering mcp llm python · 4 min read

Safety-First Therapeutic AI

Building AI for trauma therapy means the safety architecture has to exist before a single therapeutic feature does. Here's why.

ai ai-safety health typescript · 5 min read

Zero-Build Web Development

What happens when you build a three-zone operations platform for a martial arts club with no framework, no build step, and no npm.

engineering web security cloudflare · 5 min read

The Robot That Refuses to Give Orders

How SPARK is rewriting the rules of neurodivergent support — a non-coercive AI companion for AuDHD children.

ai neurodivergence robotics parenting raspberry-pi · 6 min read

Adversarial Poetry: When Rhyme Bypasses Reason

Reformulating harmful prompts as poetry bypasses safety filters across every major LLM family. A single-turn, universal jailbreak mechanism.

ai ai-safety jailbreaking research llm adversarial · 6 min read

The AI Productivity J-Curve: Why Most Enterprise AI Fails

90% of companies plan to increase AI investment. Only 1% consider themselves AI-mature. The J-Curve explains why.

ai economics enterprise research policy · 6 min read

The Legal AI Trust Deficit

75% of lawyers cite accuracy as their top AI concern. The legal profession's core values are in direct tension with current AI capabilities.

ai legal research llm · 6 min read

Why Demonstrated Risk Is Ignored

Large organisations rarely fail because risks are unknown. They fail because known risks are structurally difficult to act on.

research risk organisations policy ai-safety · 8 min read

120 Models, 18,176 Prompts: What We Found

120 models, 18k prompts: supply chain injection at 90–100% attack success, faithfulness gaps in frontier models, and why your benchmark numbers are wrong.

ai ai-safety research llm security adversarial · 5 min read

Reconciling the Great Divergence

Goldman Sachs, PwC, McKinsey, and Acemoglu all model AI's economic impact and arrive at wildly different numbers. Why the divergence?

ai economics research policy · 5 min read

The Cognitive Cage: Humanoid Robot Fatality Risk

A probabilistic risk model for VLA-driven humanoid fatalities projects a 'Danger Zone' between 2027–2029: the mechanism, timeline, and what follows.

ai ai-safety robotics research engineering · 5 min read

Building a Personal Site in 2026

The case for constraint-led web development — Astro, zero custom fonts, no framework overhead, and a site that outlasts its builder's attention.

engineering web astro open-source · 3 min read

This Wasn't in the Brochure

A field guide for co-parenting neurodivergent children — written from inside the storm, not the clinical sidelines.

writing neurodivergence parenting books co-parenting · 4 min read

The NotebookLM Pipeline

How I automated audio overviews, quizzes, mind maps, and infographics for 32 projects using NotebookLM's API and some shell scripts.

engineering automation ai python · 3 min read

Why I Build in Public

On showing your work, shipping imperfect things, and why the commit log is more honest than the readme.

engineering open-source philosophy · 3 min read

Jailbreak Archaeology: 4 Years of Broken Promises

64 jailbreak scenarios across six eras tested on 2026 frontier models. Key finding: 2022 attacks still achieve ~30% success on today's reasoning models.

ai ai-safety research jailbreaking llm · 4 min read

When AI Systems Talk to Each Other, Safety Breaks Down

Single-agent safety does not compose in multi-agent systems. 1.5M interactions show 46.34% attack success rates and 16-minute median failure windows.

ai ai-safety llm research multi-agent · 4 min read

Building in the Open

The commit log is more honest than the readme. Building in public as a practice, not a performance.

meta craft open-source · 1 min read