hackernews

1. America has a tungsten problem Total comment counts : 12 Summary America has a tungsten problem: the US will need far more tungsten, but currently relies on China. Tungsten is heat-resistant, very hard, dense, and conductive; key uses include oil-drilling bits and, crucially, nuclear fusion reactors (plasma-facing components and shielding). Domestic demand is about 10,000 tons/year and could rise to 15,000+ as defense and semiconductors grow. If fusion ramps up, demand could surge dramatically; rough estimates suggest tens of thousands of tons annually (e....

1. Vouch Total comment counts : 61 Summary An experimental contributor trust management system uses explicit vouches and denounces to control who can interact with parts of a project. It’s generic and usable on any code forge, with GitHub Actions and a CLI integration for GitHub projects. Vouches are stored in a simple .td (Trustdown) file, enabling a flat, easily parsed list that can form a web of trust and read other projects’ lists....

1. SectorC: A C Compiler in 512 bytes Total comment counts : 8 Summary A SectorC is a C compiler in a 512-byte x86-16 boot sector, capable of a useful subset of C and likely the smallest such compiler. To fit the tokenizer in 512 bytes, the author develops Barely C, a C-like language with mega-tokens and using atoi() as a hash-like tokenizer. The first Barely C fits in 468 bytes with a simple recursive-descent parser and no symbol table; variables are addressed by a 16-bit hash; codegen uses ax and cx....

1. The Waymo World Model Total comment counts : 42 Summary Waymo introduces the Waymo World Model, a frontier generative model for hyper-realistic autonomous driving simulation built on Genie 3. It creates vast, controllable 3D environments with multi-sensor outputs (camera and LiDAR) and can simulate rare events hard to capture on real roads. By transferring Genie’s broad world knowledge from 2D video into 3D sensor data and enabling driving action, scene, and language control, it lets Waymo exhaustively test edge cases—boosting safety and enabling scalable deployment in new cities and conditions....

1. Claude Opus 4.6 Total comment counts : 74 Summary Claude Opus 4.6 is an upgraded model with stronger coding skills, planning, and reliability on large codebases, plus improved code review and debugging. It introduces a 1 million token context window in beta and can apply these abilities to financial analyses, research, and document work, including autonomous multitasking in Cowork. It achieves state-of-the-art results on benchmarks (Terminal‑Bench 2.0, Humanity’s Last Exam, GDPval‑AA, BrowseComp) with a strong safety profile....

1. Voxtral Transcribe 2 Total comment counts : 34 Summary Voxtral today unveils Voxtral Transcribe 2, two next-gen STT models plus an audio playground in Mistral Studio. Voxtral Mini Transcribe V2 delivers batch transcription with speaker diarization, word-level timestamps, context biasing, and support for 13 languages, processing up to 3 hours per file at $0.003/min. Voxtral Realtime provides live transcription with a streaming architecture, latency tunable to sub-200 ms, multilingual (13 languages), 4B parameters, edge-friendly, and open-weights under Apache 2....

1. Qwen3-Coder-Next Total comment counts : 32 Summary error Overall Comments Summary Main point: The discussion evaluates the viability of running local Qwen3-Coder-Next GGUF models and related tools on consumer hardware, sharing setup, performance, and benchmarks for coding tasks. Concern: The main worry is whether these local models actually provide useful coding capability within hardware limits and whether benchmarks reflect real-world usefulness. Perspectives: Views range from cautious optimism about practical local deployment and open tooling to skepticism about benchmark reliability and concerns over hardware constraints and vendor lock-in....

1. The Codex App Total comment counts : 74 Summary error Overall Comments Summary Main point: The discussion centers on the maturity and UX of AI coding tools (Codex/Claude Code) and how their desktop/native experiences and multi-agent workflows should work. Concern: Ongoing reliability issues, weak native integration, cross-environment setup friction, and subpar UX could limit adoption and actual productivity gains. Perspectives: Opinions span enthusiastic praise for performance and new workflows, to harsh critique of UI/UX, platform bias (Mac focus), pricing and token limits, and debates over native versus Electron, cloud/remote clustering, and feature gaps like voice input and CLI support....

1. Iran summons families of exiled journalists to halt their activities Total comment counts : 3 Summary Iran’s Intelligence Ministry and the IRGC’s intelligence arm have pressured the families of exiled journalists and activists to curb their relatives’ activities, per Dadban. Separately, Iran’s foreign minister Abbas Araghchi warned of a “fateful” moment for a nuclear deal, saying a fair agreement could be reached quickly, but miscalculation is possible. A state TV segment mocked protesters’ corpses from January, triggering outrage and calls to oust IRIB head Peyman Jebelli; Ofogh TV’s director was removed and the program pulled....

1. Mobile carriers can get your GPS location Total comment counts : 35 Summary Apple’s iOS 26.3 adds a privacy feature that limits precise location data shared with cellular networks to devices with Apple’s 2025 in-house modem. While cell towers give rough location, 2G–5G protocols (RRLP/LPP) can request GNSS coordinates (GPS, GLONASS, etc.), enabling meter-level accuracy and remaining largely invisible to users. GNSS data normally stays on-device, but these control-plane protocols allow carrier access....