This is what backs the links feed on this site.

The case that you need enough depth to recognise when something is wrong. Fed into a post I wrote this week.

Argues impact depends less on years accumulated than on fresh perspective, and that candidates should be evaluated on prototypes showing taste and reasoning. Obvious reasons for my interest in this one.

The argument is about task-level value rather than job replacement, which is a more precise and more uncomfortable claim.

Simulates a 1GW datacentre losing money in 98% of outcomes. The assumptions matter more than the headline number.

Attack surface as something you can actually choose to reduce.

50 themes and openly licensed photos, aimed at stopping AI coding tools producing identical generic layouts.