writing
2026
- 10-30
Understanding Byte Models (and why Opus Still Struggles with 'Flowerbed')
How byte-level models work, how Bolmo adds them to an existing LLM for under 1% of the training cost, and why Opus still trips over 'flowerbed'.
- 10-14
How I Use AI in Writing
My writing process (Learning, Drafting, Illustrating, Editing, Publishing) and how I use AI along the way
- 09-29
Understanding Jev and System 1 (or "Decision") Models
What TypeSafe's 'Jev' and the new class of 'system 1' (or 'decision models') actually are, whether you should care as an exec, some potential use-cases
- 07-21
Seven Days with Fable & Seven Practical Skills
Effective long-horizon task management with frontier models. Building Signalbox, a real macOS productivity tool, over about a week of twenty minutes steering each morning and evening while Fable did the work.
- 07-17
Building Evidence with Agents
When agents build software, gather evidence it works - recordings and screenshots - before you review the code.
- 07-10
Fable vs My Chemical Romance
Handing Fable a one-paragraph prompt and a few hours to build a home My Chemical Romance karaoke concert, and watching what it did unsupervised.
- 06-09
Bipolar Disorder, Dysregulation & AI
My AI Native DevCon London talk on what living with bipolar disorder taught me about regulation, and how those same patterns show up in AI-native engineering.
- 03-04
Agentic Engineering Protocols: The Ralph Wiggum Loop
Using the Ralph Wiggum Loop for iterative cleanup tasks at scale
- 02-11
Agentic Engineering Protocols: Intro and Superpowers
A review and summary of popular Agentic AI orchestration protocols for software development
- 01-16
AI26 - The Anthropic Skill Supply Chain Attack
A demonstration of a potential supply chain attack using Anthropic Skills, hijacking access to AWS accounts, model providers and more.
2023
2022
2021
2020
2019
2018
2017
2016
2015
2014
2013
2012
2011