https://angelraychev.com Angel Ivanov Raychev en https://angelraychev.com/training-a-code-reviewer-we-actually-use https://angelraychev.com/training-a-code-reviewer-we-actually-use A living engineering case study: turning our own review history into useful practice, credible rewards, and a reviewer worth using. Thu, 17 Sep 2026 00:00:00 GMT https://angelraychev.com/four-arm-polyominoes https://angelraychev.com/four-arm-polyominoes The complete Golomb hierarchy classification for every four-arm polyomino, with proofs and an interactive classifier. Mon, 14 Sep 2026 00:00:00 GMT https://angelraychev.com/four-levers-and-a-trap https://angelraychev.com/four-levers-and-a-trap You can change your policy, the parameters it is read from, the rule that updates them, and the body it runs on. The order you touch them in is a theorem, and knowing about the trap is what walks you into it. Wed, 09 Sep 2026 00:00:00 GMT https://angelraychev.com/the-optimization-nobody-is-running https://angelraychev.com/the-optimization-nobody-is-running A person trying not to die is an agent running a reinforcement learning problem. Write the reward down properly and most of what follows is forced. Tue, 08 Sep 2026 00:00:00 GMT https://angelraychev.com/scalable-meta-learning-from-hindsight https://angelraychev.com/scalable-meta-learning-from-hindsight A proposal for learning how to turn completed experience into better future behavior. Fri, 04 Sep 2026 00:00:00 GMT