TSB

(LLM written article) Everyone has their personal benchmarks for testing the latest AI models. One question I had in mind: can agentic AI reproduce itself? I ran this experiment on Opus 4.5 in January 2026, providing only vague prompts. To my amazement, it succeeded with a fully verified Verilog implementation of a custom processor architecture including assembler written firmware for LLM inference.

Building wireframe polyhedra made from LED filaments, using graph theory to devise geometry and driving strategies.

Combining a deep-depthwise CNN architecture with variable quantization in BitNetMCU achieves state-of-the-art MNIST accuracy on a low-end 32-bit microcontroller with 4 kB RAM and 16 kB flash.

(Guest article on the Nous Research blog) Anecdotal evidence suggests open weight models produce significantly more tokens for similar tasks than closed weight models. This report systematically investigates these observations. We confirm this trend to be generally true, but observe significant differences depending on problem domain.

Todays candles have been optimized not to flicker. But it turns out when we bundle three of them together, the resulting triplet will start to naturally oscillate. Amazingly, the frequency is rather stable at ~9.9 Hz as it mainly depends on gravity and diameter of the flame. We detect the oscillation with a suspended wire and divide it down to 1 Hz.