I found a collection documenting various typographical errors stamped into the concrete of San Francisco's sidewalks. It captures permanent mistakes made by city workers, ranging from simple misspellings to completely inverted letters.
Explore Malcolm Gladwell's argument that limited information can improve decision-making through heuristics. A study of students shows how recognizing only one city name can lead to more accurate population estimates than having access to more data.
I look at Kamato Hongo's 115th birthday and her unique sleep-wake cycle, where she sleeps for two days and stays awake for two days. This unusual rhythm highlights the mysterious habits of extreme longevity in Japan.
I used demucs and ffmpeg for audio, tested Astral’s ty type checker, and updated my TTS cost analysis. I also explore AI personhood, model self-correction, and bizarre self-driving car mishaps involving the moon and wet cement.
I explored ChatGPT's export limits, Zettelkasten note-taking, and AI agent architectures from the AI Engineering Summit. I also noted Tyler Cowen’s LLM writing workflows, new OCR models, and the 'Diffusion LLM' approach to fixing hallucinations.
I leverage the implicit guarantees in puzzles and exams—such as the certainty of a unique solution—to solve problems using simple trial and error. I demonstrate how identifying even one valid case can reveal universal answers.