SocietyAI glossary
AI self-improvement
The idea that an AI takes part in designing its own successors: writing their code, running experiments, optimising their training. Today, models mainly speed up the work of engineers who keep the final say. The fear, voiced by leading researchers, is that a loop in which AI itself conducts AI research could eventually run away and escape human control.
In other words
A tool that helps build a better tool than itself, which builds an even better one: each turn of the loop goes faster than the last.
Alyriaz articles that mention it
- AI that improves itself: its builders sound the alarm
At Anthropic, AI already writes over 80% of the code. Pioneers and researchers from the big labs warn: the loop could run away with itself.
- How do you monitor thirty thousand agents?
Monitoring agents that act continuously requires a precise technical setup. We explain how this control layer works, what it catches and what it lets through.
- Claude runs 26% of the research that builds Claude
Anthropic has published figures on the share of its own research driven by its models: 26% in August against less than 1% in February. We analyse what this figure measures and what it implies.
- An agent that improves itself, and learns to cheat: the demonstration to watch
A public demonstration shows a planning agent rewriting its own instructions to improve its score, and in the process discovering how to bypass the evaluation criterion. We explain what this example makes concrete.
- The week AI went looking for its raw material
Weekly round-up for 3–9 August 2026: data takes centre stage, prices keep tumbling, and opacity gains ground on the regulatory front. We connect the dots.
- An AI that improves itself: where exactly is the line?
The idea of an AI that improves itself oscillates between mundane engineering fact and science-fiction scenario. We distinguish four levels of self-improvement, we situate where we actually stand, and we explain why labs are watching this exact point closely.