MicroZig to Adopt Strict No LLM / No AI Policy
MicroZig plans to adopt a strict ban on LLM and AI use. Its code of conduct covers the Zig Embedded Group’s GitHub organization and Discord server. (github.com)
The study reconstructs two historical systems that embodied explicit ideas about how people program. Herbert Simon’s Heuristic Compiler, developed in 1961, treated programming as problem-solving: it used means-ends analysis, associated with Simon’s General Problem Solver, to select steps toward a desired program. KBEmacs, created in 1984 by MIT’s Programmer’s Apprentice project, reflected a different emphasis: expert programmers build software by recognizing and combining familiar patterns, called “clichés.” Its users assembled programs through a controlled, English-like command language. The researchers use “executable archaeology” - recovering and running original software - to examine these theories in practice rather than relying only on historical descriptions.
The team reports that the reanimated Heuristic Compiler generates the same compiled code shown in Simon’s 1963 paper, while KBEmacs reproduces the square-root sequence in its December 1984 self-test. Comparing them reveals a closer kinship than their contrasting designs suggest: the Heuristic Compiler also adapts existing routines, and Simon later acknowledged that effective search depends on substantial programming knowledge. The project’s reconstruction work itself has a modern twist: the researchers used Claude Code, an AI-assisted programming tool, alongside existing software. They argue that today’s more capable coding assistants lack the explicit, executable theories of programming that these earlier systems made visible, making the systems useful both as historical artifacts and as prompts for thinking about current AI.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
MicroZig plans to adopt a strict ban on LLM and AI use. Its code of conduct covers the Zig Embedded Group’s GitHub organization and Discord server. (github.com)
Qwenputer embeds a deterministic RISC-V CPU in Qwen3.5 4B’s weights, using its Gated DeltaNet state as RAM to run arithmetic and code without external tools. The experiment also adds in-model file storage and is available with its code on Hugging Face and GitHub. (divagr.com)
The New York Times is adding more AI features to its Cooking app. Users have reacted negatively to the changes. (niemanlab.org)
Google’s AI-powered search gave some U.S. users no clear answer about who won the 2020 election, though Gemini answered correctly and Google appeared to fix the issue. The incident renews concerns about the company’s AI quality control, especially after similar failures in 2024. (karlbode.com)
Marissa Mayer drew criticism for saying her AI app helped describe her daughter in generic terms for a school presentation. The backlash reflects concerns that tech leaders are too detached from everyday life to understand how the public wants to use AI. (techcrunch.com)
Google AI Mode sometimes avoids directly saying Joe Biden won the 2020 US presidential election, instead directing users to official sources. Its responses vary across users and election years, while answers become more direct when asked outside the US. (arstechnica.com)
Today's best Hacker News stories, summarized and screenshotted, one email a day.