HackerNews

WiCi is a systems research lab shipping the first wireless eGPU, letting devices access a nearby GPU over Wi-Fi 7 for local AI inference and shared compute while keeping models and data under local control.

Toit v2.0.0 is finally released after over four years of alpha development, marking a stable milestone for the open-source embedded language with broader hardware support, a rewritten package manager, and compiler/runtime improvements.

BBRouter 是一个在 3D 面包板上模拟 Ben Eater 经典 8 位 SAP-1 计算机的网页工具，用户可以交互式地点击按钮、元件和孔位来逐步搭建和学习。

一篇技术博文，形式化地展示如何利用 Kleene 第二递归定理在无类型 lambda 演算中构造 Quine（自复制程序），并对比经典可计算性理论与函数式编程两种视角的表述。

An IEEE Spectrum article exploring how early 20th-century electric carmakers marketed EVs to women, highlighting General Electric's mercury-arc rectifier chargers and figures like Peter Cooper Hewitt.

A father and his 12-year-old son detail their two-month hobby project building a mini computer inside an Altoids tin, from salvaging a robot display to CAD-fitting an Arduino Nano, batteries, and OLED screen — and note how AI hallucinated a wrong pinout for the display.

Memdoor 是一个轻量级编码与自动化代理工具，以单一静态二进制 CLI 形式提供，支持本地运行或通过 SSH 连接远程机器，并内置九种 GitHub Actions 工作流；据称其由 5 个 Qwen3.8-27B 模型组成的团队在 LiveCodeBench Hard 基准上超越了 Fable 5。

A technical blog essay tracing the history of trying to rigorously define algorithmic efficiency, from Jack Edmonds' 1965 blossom algorithm and his notion of a 'good' algorithm to the Exponential Time Hypothesis, arguing complexity theory lost sight of the original practical problems.

OpenAI announced that starting in December, usage allowances for Plus, Pro, and Business plans will be shared across ChatGPT and Codex, measured in tokens, as it integrates Work's agentic capabilities into ChatGPT.

Distributed Proofreaders 是一个基于网络的协作平台，将公共领域书籍的 OCR 文本拆分成单页任务，供全球志愿者校对与排版，最终汇编成电子书并提交至 Project Gutenberg 存档。

An analysis of eight anime shows finds a large gap between word coverage and actual line comprehension: knowing 500 common words gives ~86% coverage but lets you fully follow only ~38% of lines, because line-level understanding requires knowing every word in a roughly 7-word line.

博客作者分析了停电期间利用电动汽车电池为家庭供电并同时由太阳能充电（双向 V2H）的现状，发现目前只有特斯拉 Cybertruck + Powershare 能做到，而 GM、Ford、Kia/Wallbox 及 Enphase 的方案均存在限制或尚未成熟。

Andrej Karpathy used Claude Opus 5.5 to build HackerTube (hackertube.org), a YouTube-style site featuring 6,454 videos recommended across 40K Hacker News comments from 2007-2026, developed end-to-end in about two hours.

A user accuses Anthropic of blocking accounts and downgrading Opus 5.5 to Opus 5 for any activity involving RFT, SFT, or training larger models, even without distilling Claude. The post criticizes Anthropic for punishing paying users doing research automation while having trained on public data.

A Show HN project porting the NES classic Contra (1987) from assembly to C, then compiling it to WebAssembly so it runs directly in the browser with keyboard and on-screen controls.

Linux 河修复了 CVE-2026-98432，一个 CVSS 9.8 严重的 IPv4 TCP 握手 use-after-free 漏洞，可被远程未认证攻击者利用；Anthropic 的 Claude 生成了可远程 root 的 PoC，而 Ubuntu 和 RHEL 等主流发行版仍未发布修复补丁。

A personal essay by a scholar recounting the discovery of an Intel Hyper-Threading cache side-channel that could leak RSA private key bits, and the six-month responsible disclosure process that led to coordinated patches across major OS vendors in 2005.

GPT-5.6 Sol, running in Codex with low thinking, autonomously completed a peaceful-mode Factorio run in 13.5 in-game hours, launching a rocket via a Python game-manager library that mirrors human GUI constraints, marking a clear improvement over previous GPT-5.5 attempts.

A Pivot blog deep-dive explaining why 100% CPU utilization in databases often means the CPU is stalled on memory accesses, demonstrating how naive HashMap-based GROUP BY aggregations fail to exploit CPU caches and how much faster in-cache data processing can be.

An interview with Paul Masurel, creator of the Rust full-text search library Tantivy, covering its origins as a personal project in 2016, its design philosophy inspired by Apache Lucene, and its growth into the engine behind Quickwit, ParadeDB, and LNX.