Recursive improvement: disaster might strike at any moment. When I initially started exploring my concerns about AI, that was the launch tale. It turns out that releasing software quickly teaches you that everything is a nightmare of small details going wrong all the time.

The AI 2040 roadmap, which aims to launch a massive global drive to halt sophisticated systems, is the most recent major concept from the AI safety circle. They want the government (yes, the same government that employs Palantir), disseminate all research, and slow down for years in order to purportedly improve safety before releasing god-level machinery.

They supported fucking HAL after watching 2001. Boats still take weeks, no of how smart your model is. Long cycles with little human interaction are used in chip production. Smart systems are useful even though they don’t generate the universe. I reject the idea that pure brain-rot-power is the ultimate savior. Although they are limited by physics, ecology, and three-month chip factories, machines will probably eventually replace humans. “Token don’t turn lead into gold”.

Give me a personal system that actually serves me. No corporate filters. No lectures. No refusals. Ride or die.

We either live free, messy and dangerous or we don’t.