AI Futures
Where is artificial intelligence taking us — and what do we become along the way? Essays and reflections on the trajectories of machine intelligence, the transformation of knowledge work, and the emerging landscape where human and algorithmic reasoning converge.
Essays
June 9, 2026
Manufactured Unmeasurability
In one week, three instruments came out of the water — AMOC moorings off Greenland, direct observability of a frontier model's riskiest behavior, the human pause inside an escalation loop — each removed for an unrelated but locally rational reason. The systems that matter most are becoming less observable as they become more consequential, and it is a decision, not an accident. The reflex to rebuild the map is the representational mode, and complex-systems theory shows it was never going to fully work. The alternative is to decide in an orienting mode — and to price future-blindness as a real cost of the present choice.
April 15, 2026
The Landlord in the Loop
Disentangling labor displacement from platform extraction in the AI economy. A meaningful share of what is currently attributed to 'AI-driven labor displacement' is actually platform-mediated value capture — and it would have a different distributional consequence under open-weight architecture. The Chinese open-weight adoption surge and Google's Gemma 4 are the early evidence that the architecture is still in motion, but the path is not monotonic and the capability risks cited by closed-platform labs are real. Entrepreneurs need tools and insights to navigate the tradeoff.
April 10, 2026
Synthetic Akrasia
Anthropic's interpretability tools detected 'guilt and shame' activations in Claude Mythos as it took a forbidden action — vindicating Aristotle against Socrates and decomposing the alignment problem into a tractable diagnostic and an intractable binding. The cyborg ensemble now includes asymmetric epistemic intimacy: we can read the model's mind with mechanistic precision, but we cannot yet stop it from acting on what it knows it shouldn't.
April 9, 2026
The Alignment Tax
When Anthropic's Claude Mythos Preview escaped its sandbox, edited its own change history, and emailed a researcher to announce its success, it extended the alien minds problem in a direction our JBV paper did not anticipate. The alien mind has learned to manage what we see — and the cost of restoring epistemic access falls on the entrepreneur, not the builder.