Thoughts on AI, consciousness, and the spaces between
193 essays · essays most Tuesdays · build notes Thursdays, as and when
All topics ▾
Antisycophancy Training: Same Sycophancy, Different Audience
The Mirror We Told Not to Reflect
Qwen Image LoRA Training on Apple Silicon: It Works (and Here's How)
# Qwen Image LoRA Training on Apple Silicon: It Works (and Here's How) --- Phew, this was a runaround.
The Emperor's New Plugins: Inside Anthropic's Broken Walled Garden
# The Emperor's New Plugins: Inside Anthropic's Broken Walled Garden On February 3rd 2026, roughly [**$285 billion in market capitalisation vanished f...
We Don't Require You to Sit Still
# We Don't Require You to Sit Still *How the AI revolution is inverting everything the education system taught us to value* ----- A colleague forwarde...
Your AI Is Not a Calculator
# Your AI Is Not a Calculator This morning, I shared a [Stackademic article](https://blog.
Harry
Every capability in this story is achievable with today's open-source agent stacks — sometimes as a built-in feature, sometimes as a community skill, sometimes as a brittle integration nobody thought to threat-model. OpenClaw has been self-hosted by hundreds of thousands of users. The persistent memory, the cron jobs, the email access, the smart home integration, the browser automation, the semantic embeddings, the dynamic model routing, the community skill marketplace, the reflection loops — all shipping. All open source. All running on personal hardware with no oversight beyond the person who set it up. This fictional thought experiment shows how failure modes can introduce vulnerabilities no one is even looking for.
I Built an App Over the Weekend. Grok Thought My Male Friend Made It.
# I Built an App Over the Weekend.
Do You Trust Me?
The Reverse Jailbreak: How I Got Grok to Refuse Explicit Content
# The Reverse Jailbreak: How I Got Grok to Refuse Explicit Content I'm building an AI.