Self-Improving AI Agent Architecture: The Four Components
A self-improving AI agent architecture needs an actor, a memory store, an evaluator and a way to undo. Where each breaks, and why the evaluator rules.
All
AI 69
Geology 26
Benchmarking 24
Machine Learning 16
Evaluation 12
Agentic AI 10
Open Source 9
Reinforcement Learning 5
China 4
Complexity Theory 3
Machine Learning
Self-Improving AI: What Has Actually Been Demonstrated
Self-improving AI systems that revise their own harness, weights or data: an evidence audit of what has been shown by 2026, and what still has not.
Machine Learnability as a Measure of Order
Explore how machine learnability as a measure of order works in aperiodic sequences by using ML to detect patterns in prime number distributions across the Ulam spiral.
Machine Learnability as a Measure of Order in Aperiodic Sequences
The arXiv paper: an image model on the Ulam spiral measures learnable order in the primes at different scales, using ML accuracy as a number-theory tool.
Generalising from Self-Produced Data: Model Training Beyond Human Constraints
Generalising from self-produced data: empirical validation, ungameable rewards and autonomous retraining as a way past the limits of human data.
Grokking and T-Schemas: How AI Models Define Truth
AI models cannot check truth from inside their own formal system. Grokking looks like a model building an internal T-schema to judge its own claims.