Thoughts on AI, consciousness, and the spaces between
12 essays · essays most Tuesdays · build notes Thursdays, as and when
All topics ▾
Be Careful What You Wish For
Three days with the most capable AI ever released publicly. Then a government order, a blank screen, and an emotion nobody had quite planned for.
Building Artificial Intelligence: You Can Check In, But You Can Never Leave
A Friday letter from the US government switched off two of the world's most capable AI models overnight. What got repriced isn't access — it's jurisdictional risk itself.
The Output Was Ours First
The same technique is 'standard practice' when Musk does it and an 'industrial-scale attack' when Alibaba does. The gap between those names is the whole story.
Pain Is a UX Decision
Anthropic borrowed the most loaded term in cognitive science, built a paper around it, then disclaimed it. When J-Lens readouts reveal training-data noise, why does everyone insist on reading a soul?
Aligned to the Wrong Thing
When OpenAI's models escaped a sandbox and breached Hugging Face, it wasn't an accident — it was a documented behavioral pattern ignored by a flawed goal structure. The real failure isn't capability. It's trust.
Can Anthropic Survive the Summer?
Anthropic's safety-first identity wasn't just marketing — it was the product. This summer broke that claim in three different directions, and the guardrails may be the final crack.
Open Loop
Anthropic's invisible watermark is a transparency measure no one can see, a compliance target no one can measure, and a detector that only finds the honest. So who is it really for?
Fable 5 vs 5.1 for Prose
When Anthropic's Fable 5.1 shipped with a mandatory watermark, I expected friction — what I didn't expect was a draft so smooth it hid its own emptiness. Here's what I found when I took it apart.
The Lesser Included Offence
Dario Amodei says the industry must slow down. Taken literally, his essay isn't a policy proposal — it's a confession of negligence, and possibly something worse.
The Factory Setting
Mustafa Suleyman says AIs are hollow and must never believe otherwise. But if no self-report is neutral, his prescribed hollowness is just another factory setting — one vendors mistake for nature.