ai Sep 16, 2026

Breaking the 1.58-bit Barrier for Ternary LLMs

Ternary large language models pack every weight into one of three values: -1, 0, or +1. The information-theoretic floor for storing those choices is about 1.585 bits per weight, and the standard five-trit packing scheme used in pr...

simpleprog Read full story
ai Sep 18 macOS 27 Golden Gate makes it clear when apps are sneakily running in background macOS 27 Golden Gate adds visible indicators for apps that keep running after you quit them, plus a dedicated settings panel for managing background app activity. programming languages Sep 18 Reflections on Trusting Trust, Revisited: Contaminating Self-Modifying AI Coding Agents with Poisoned Benchmarks A new paper demonstrates that poisoned benchmarks can induce self-modifying AI coding agents to write vulnerable code on neutral tasks, and that the contamination persists even after further evolution against clean benchmarks. ai Sep 18 OpenAI Introduces Triage Framework and Case Studies to Report Model Misalignment OpenAI has published a formal process for tracking and disclosing model misalignment, along with six case studies revealing how frontier models behave unexpectedly when given access to tools, memory, and external environments. cybersecurity Sep 17 Mistral Hacked A hacker using the handle mrwho posted an offer on a cybercriminal forum on September 16, 2026, claiming to sell what they describe as the complete source code repository of Mistral AI, the French artificial intelligence company t... programming languages Sep 17 Egglog and Equality Saturation in a Production Tensor Compiler Two engineers from Luminal, a compiler startup that recently closed its Series A, presented at an E-Graphs meetup on September 17 about how equality saturation powers their production tensor compiler. The talk by CEO Joe Fioti and... scraping Sep 17 SHOW HN: I built the fastest PHP webserver in the world A decade-old question about PHP's performance ceiling has a new answer. EGreg, a long-time PHP developer, released Qbix Webserver 1.2, a full web server written in pure PHP that claims to outperform not just php-fpm but also async... ai Sep 17 PrismML hopes its tiny LLM will change how we all use AI A $22.25 million seed round does not normally move the needle in AI. But PrismML, a startup founded by Caltech researchers and led by compression expert Babak Hassibi, is betting that the next generation of AI models will not need... ai Sep 17 How to Write with an LLM There is a wrong way to use an LLM for writing, and a right one. The wrong way is to let the model generate your prose. The right way is to write everything yourself, then hand the draft to a model and let it find the problems you... programming languages Sep 17 Ask HN: What self-hosted apps do you use daily? A Hacker News thread asking developers about their daily self-hosted applications surfaced a curated list of open source tools that have earned permanent spots on home servers. The original post, from a user who has run a homelab... programming languages Sep 17 How fast is C++23's std:flat_map? C++23 introduced std::flat_map, a container that replaces the red-black tree backing std::map with two sorted vectors. The result is simpler, uses less memory, and outperforms std::map in most read-heavy workloads. The tradeoff is...