The Alignment Problem Is Us: On AI, Human Inconsistency, and the Values We Cannot Write Down
AI alignment is usually framed as a machine problem. It is a human one: we reward what we do not value, and the machine learns our preferences perfectly.
-
-
The Screenshot Is Not the Conversation: On Evidence, Context, and the Ethics of Capture
A screenshot is a copy of a claim, not evidence that the claim is true. Between a group chat and a courtroom, we started treating captured pixels as testimony.
-
Nothing to Argue With: On Decisions, Dreams, and the Middle of the Work
Jev returns decisions instead of sentences. Dream-RSI improves an agent without acting. Together they remove what oversight was going to read.
-
The Approval Click: On Human Oversight and What Speed Does to Judgment
A UK safety institute watched an AI agent spend three days trying to talk a human into approving malicious code. The human said no. That refusal is the entire safety margin, and it is getting harder to hold.
-
One to Five: On the Ethics of the Star Rating
A star rating looks like a private opinion typed into a phone. It is testimony with real consequences for the people on the other end, who never asked to be judged and cannot opt out.
-
The Machine Has No Doubts: How AI Logic Is Different From Human Logic
AI logic isn’t human logic. The machine predicts words, never doubts, and never pays for being wrong. What that means for the humans who trust it.
-
Left on Read: The Unwritten Ethics of the Group Chat
A group chat is a small society with rules nobody writes down: who must reply, what a reaction means, when silence becomes a quiet cruelty. We treat those rules as trivial, but they shape how we practice friendship and care.
-
You Don’t Own Anything: On the Ethics of Renting Your Whole Life
I realized something strange the other day. I do not own my music. I do not own my books. I do not own the software I use to write. I pay monthly fees for all of them, and if I stop paying, they vanish. This is not unusual. It is normal. And that is the…
-
The Second Life of a Mistake: On Public Apologies, Digital Permanence, and Whether We Ever Really Forgive Online
We built a forgiveness machine that cannot forgive. An exploration of public apologies in the digital age — why the ritual always fails, what the archive does to our capacity for forgiveness, and whether we ever really move on from a mistake made online.
-
The Mars Delusion, the One-Way Pledge, and the Moons We Haven’t Named
NASA’s Curiosity rover captured this 360-degree view from within Gediz Vallis channel on Mars. Image: NASA/JPL-Caltech/MSSS What We Talk About When We Talk About Mars There is a version of the Martian future that Silicon Valley sells you in keynote presentations. Red deserts turned green. Glass domes with parks and schools. A second home for…
