HN Top 10 — November 03, 2026
Today’s Top 10 on Hacker News
1. Astra can read and write weights to improve Qwen
⭐ 2749 💬 834 👤 dbreunig 🔗 Discuss on HN
The article describes a technique called Weight-Space Programming, in which large language models directly inspect and modify neural network weights without gradient-based training. It reports that an LLM can infer a network’s intended task from raw weights, generate small game-playing networks, and improve a Qwen model’s math performance by making tiny weight edits. These edits reportedly boost math benchmark scores substantially and also improve instruction-following.
2. Nvidia – open-source auto-ontology
⭐ 104 💬 32 👤 campervans 🔗 Discuss on HN
NVIDIA Auto Ontology is an open-source library that builds a semantic context layer for AI agents from enterprise metadata. It generates and reconciles ontologies across systems, grounds metrics and entities with trusted business meaning, validates generated queries against certified definitions, and exposes the context to any agent harness via MCP, CLI, and API. It is Apache-2.0 licensed, uses NVIDIA NIM models for inference, and currently does not accept external contributions.
3. Toit v2.0.0
⭐ 117 💬 37 👤 floitsch 🔗 Discuss on HN
Toit v2.0.0 is the first major release of the open-source Toit language after years of alpha development, following the shutdown of its original company and continued community maintenance. The release finalizes a major breaking change that lets peripheral APIs accept pin numbers directly and adds broad hardware support, a unified command, a rewritten package manager, mixins, firmware envelopes, and compiler and runtime improvements. Future releases will continue adding features such as a formatter, compiler resolution improvements, and support for more architectures.
4. DiscImageStudio: burn visible artwork onto CDs, DVDs and Blu-ray discs
⭐ 253 💬 52 👤 bpierre 🔗 Discuss on HN
DiscImageStudio is a tool for creating optical disc images with visible artwork burned directly onto the data layer of CDs, DVDs, and Blu-ray discs. It likely lets users add custom graphics or labels to physical media during the disc burning process.
5. Katamari Damacy but it’s The Met collection
⭐ 25 💬 5 👤 thebigship 🔗 Discuss on HN
Katamari Met is an interactive web project that lets players roll up real objects from the Metropolitan Museum of Art’s Open Access collection at their actual sizes. Starting with small items like coins and thimbles, the experience builds toward larger pieces such as the Temple of Dendur.
6. Sava Trust (YC F25) Is Hiring a Founding Engineer in NYC
⭐ 1 💬 0 👤 nimz 🔗 Discuss on HN
Sava Trust is a Y Combinator-backed startup seeking a founding engineer in New York City to help build a modern trust company. The company is creating a full-stack, chartered trust entity that manages family wealth and trust assets while executing the wishes of families.
7. Show HN: Hacker Times, a New York Times-inspired version of Hacker News
⭐ 312 💬 97 👤 polpenn 🔗 Discuss on HN
Hacker Times is a New York Times-inspired presentation of Hacker News that turns community links into a curated, newspaper-style front page. It highlights technology topics such as AI-assisted software development, neural-network editing, privacy-conscious hardware, web archiving, and open-source projects.
8. European RISC-V Industrial Computing Initiative
⭐ 136 💬 72 👤 ymz5 🔗 Discuss on HN
The European RISC-V Industrial Computing Initiative is developing an open, non-commercial reference stack for industrial computers based on RVA23 RISC-V hardware. It defines and validates the interfaces between firmware, boot software, a partitioning hypervisor, real-time control, and Linux, so that compatible systems can run isolated workloads on a single SoC. ERICI aims to close the gap between available RISC-V silicon and the reliability, field connectivity, and long-term independence needed for industrial use.
9. Show HN: Subvocal – Talk to your computer without making noise
⭐ 722 💬 345 👤 aarushyadav
Subvocal is a Show HN project — a pair of smart glasses that lets users issue voice commands without audible sound by capturing subvocal speech. The discussion highlights excitement about hands-free and accessibility use cases, along with questions about where the signal processing happens (on-glasses vs. app), privacy concerns, and whether the tech could work with existing prescription frames.
10. Vathographs: Photographic depth maps as creative expression
⭐ 28 💬 6 👤 hankbond 🔗 Discuss on HN
Vathography is a creative practice that transforms ordinary photographs into depth-based artworks by using AI to infer the relative distance of every point in a scene. Instead of relying on light and color, it treats volume as the primary element, allowing artists to remap focus and emphasize different distances through tone and contrast. This turns a normally hidden technical depth map into a new way of seeing and composing photographic subjects.