I Tested AI Context Compaction. The Clever Fork Lost.
I built a reasoning-continuity benchmark for AI context compaction. A cache-preserving prefix fork worked, then lost on quality, cost, and latency.
rico.codes is a personal blog by software engineer Rico Kahler
I built a reasoning-continuity benchmark for AI context compaction. A cache-preserving prefix fork worked, then lost on quality, cost, and latency.
Loop engineering is not about abandoning prompts. It is about moving repeated steering work into verifiable systems so attention can stay on judgment, review, and taste.
A personal essay on where AI agents pay off: system design, benchmark harnesses, parallel work, token economics, and great foundations.
MAGA explained through reactionary idealism
AI feels profound when it collapses the cost of executing intent. That is not the same as machines having human will; it is delegated agency, and it is why UCA feels close for knowledge work with evaluable outcomes.
Is AGI inevitable, or is there a "ceiling" to artificial intelligence? This article explores Universal Computable Automation (UCA)—a theory proposing that while AI execution is infinite, human judgment remains non-computable. Discover why the Turing limit, the Anthropic Principle, and simulation theory suggest a future where humans define the constraints, not the machines. A pragmatic look at why…
Want to know why there's a new JavaScript framework every week? It's not all that bad actually.
'Performance' is a loaded word in the frontend/React world and hard to know exactly what people mean. This article explains the difference between web performance and re-render performance.
On Sunday, January 16th, I made wordlepdf.com, a remake of Wordle… expect it uses PDFs.
Here's my thoughts on the good as well as the bad of React Hooks
I watched all of React Conf 2021. Here's everything you should know.
React has been focusing more and more on server-related features prompting familiar the question: Is React a framework or a library?