Published on June 18, 2026 at 18:01 CEST (UTC+2)
Swiss parliament lifts ban on new nuclear power plants (123 points by leonidasrup)
Swiss parliament lifts ban on new nuclear power plants
The Swiss parliament voted during the summer 2026 session to lift a ban on constructing new nuclear power plants, reversing a post-Fukushima decision. The move aims to address energy security and federal finances, with debate covering the 13th AHV pension and armed forces. This policy change could influence future energy infrastructure for data centers and high-tech industries.
Emacs 31 Is Around the Corner: The Changes I'm Daily Driving (250 points by frou_dh)
Emacs 31 Is Around the Corner: The Changes I'm Daily Driving
The author describes new features in the upcoming Emacs 31 release, particularly a “just works” tree-sitter integration for syntax highlighting and code analysis without external packages. Other improvements are being tested from the emacs-31 branch, emphasizing native support for modern parsing. The article reflects ongoing evolution in developer tools, especially around language-agnostic code understanding.
Ubiquiti: Enterprise NAS, Built on ZFS (74 points by ksec)
Ubiquiti: Enterprise NAS, Built on ZFS
Ubiquiti introduces the ENAS, an enterprise storage appliance using ZFS, ARM Neoverse N2 cores, 64GB ECC memory, and optional NVMe L2ARC caching. It supports up to 16 drive bays and over a petabyte of raw storage with no firmware restrictions on drives. The system targets organizations needing high-performance local storage without proprietary licensing, appealing to AI/ML data pipelines.
What was nice about the UI of Windows 2000 (45 points by npongratz)
What was nice about the UI of Windows 2000
The author nostalgically reviews the Windows 2000 interface, highlighting clean design, clear font rendering, skeuomorphic buttons, and intuitive start menu behavior. They contrast it with modern “flat” UIs, arguing that visual affordances made the system easier to learn. The piece is a retrospective on UX principles rather than a technical analysis.
I found 10k GitHub repositories distributing Trojan malware (211 points by theorchid)
I found 10k GitHub repositories distributing Trojan malware
The author discovered a large-scale malware campaign on GitHub where repositories impersonated popular projects by copying commits and then adding malicious zip archives. Over 10,000 repositories shared a common pattern, allowing automated detection. The post details the investigation and raises alarms about supply chain attacks through open-source platforms.
A website that lists websites to submit your website to (24 points by azeemkafridi)
A website that lists websites to submit your website to
Submission.Directory is a hand-picked catalog of platforms for promoting websites, including launch platforms, startup directories, and AI tool listings. It provides authority scores (DR) and details on backlinks, catering to SEO and visibility. The site targets founders and marketers looking to gain traction and backlinks.
Launch HN: TesterArmy (YC P26) – Agents that test web and mobile apps (22 points by okwasniewski)
Launch HN: TesterArmy – Agents that test web and mobile apps
TesterArmy uses AI agents to automatically test web and mobile applications by describing tests in plain English. It integrates with CI/CD, Slack, and GitHub, producing screenshots, recordings, and bug reports. The service aims to catch broken flows without requiring test scripts or SDKs, streamlining QA for modern teams.
Hospitals and universities repurposing drugs at 90% lower cost (156 points by giuliomagnifico)
Hospitals and universities repurposing drugs at 90% lower cost
Research from King’s College London reveals that universities and hospitals conduct late-stage clinical trials for repurposing generic drugs at less than 10% of pharmaceutical industry costs. This “hidden” innovation system operates outside patents, using lower barriers for expertise, risk, and capital. Examples include repurposing a cancer drug for blindness and an anti-inflammatory for COVID-19.
Has W Social switched to closed source? (113 points by nemoniac)
Has W Social switched to closed source?
The article investigates W Social, a Bluesky fork marketed as a European alternative to X with identity verification and European data hosting. The author alleges discrepancies between its public claims and actual practices, including possible closed-source code and plans to train European AI models on user data. The piece raises concerns about digital sovereignty and transparency.
Dwarf Fortress in the Browser (25 points by andre9317)
Dwarf Fortress in the Browser
Remote-DF is a Docker-based solution that runs Dwarf Fortress (Classic or Steam) on a remote x86-64 Linux host and streams it to a browser via noVNC with audio. It provides full native speed gaming without local installation. The project demonstrates practical remote desktop streaming for resource-intensive applications.
Tree-sitter adoption in Emacs signals a shift toward language-agnostic parsing for AI-assisted coding
Emacs 31’s native tree-sitter integration means editors can leverage robust, incremental parsers for syntax highlighting, code navigation, and refactoring without per-language hacks. For AI/ML, tree-sitter provides structured code representations that can feed into code completion models and static analysis tools. Why it matters: As LLMs like GitHub Copilot rely on accurate code context, tree-sitter enables faster, more reliable semantic parsing directly in editors. Actionable: AI coding tools should consider tree-sitter grammars as a lightweight alternative to AST-based approaches for real-time assistance.
High-performance local storage (ZFS + ARM Neoverse) aligns with growing on-premise AI/ML data pipelines
Ubiquiti’s ENAS combines ZFS data protection, NVMe caching, and ARM Neoverse N2 compute to deliver enterprise storage without proprietary licensing. This reflects a trend where organizations keep AI training and inference data locally for latency, compliance, and cost control. Why it matters: Cloud egress costs and data sovereignty regulations push AI teams toward hybrid or on-prem solutions. Implications: Expect more “AI-optimized” storage appliances with integrated compute and ZFS-like features to support large model datasets.
Large-scale malware on GitHub demands AI-driven supply chain security
The discovery of 10,000+ repositories distributing trojans via commit poisoning shows how open-source ecosystems are targeted at scale. Manual detection is infeasible; AI/ML models that analyze commit patterns, code diffs, and behavioral anomalies are essential. Why it matters: As AI models increasingly train on open-source code, poisoned data can compromise downstream models. Takeaway: Platforms like GitHub should deploy real-time ML anomaly detection for suspicious repository clones and URL injections.
AI agents for automated testing (TesterArmy) represent a maturation of LLM-based QA
TesterArmy uses natural language to define test scenarios, then AI agents navigate apps, capture screenshots, and log bugs. This aligns with the industry trend of “AI QA engineers” that reduce human effort in regression testing. Why it matters: Combined with CI/CD, these agents enable continuous testing with minimal script maintenance. Actionable: Teams should evaluate agent-based testing for high-touch flows (e.g., e-commerce checkout) while monitoring for false positives and coverage gaps.
Drug repurposing at 90% lower cost highlights an underserved area for AI-driven discovery
The King’s College research shows that academic/hospital-led trials repurpose generic drugs far cheaper than industry. AI can accelerate identifying candidate drugs by mining biomedical literature, electronic health records, and molecular data. Why it matters: The “hidden” innovation system lacks capital but has expertise; AI can lower the expertise barrier further by automating hypothesis generation. Implications: Startups should focus on AI models that predict repurposing opportunities, then partner with academic hospitals for low-cost trials.
Social media platforms like W Social reveal tensions between AI training data and open-source ethics
W Social’s promise of European digital sovereignty conflicts with alleged plans to train AI models on user data behind closed code. This mirrors broader debates about data rights versus AI progress. Why it matters: AI models require massive text data, and social platforms are prime sources. Takeaway: Expect more regulation around “AI training consent” and calls for transparent data sourcing. Developers should factor in data provenance when building models from social media.
Remote streaming for heavy applications (Dwarf Fortress in browser) parallels remote AI inference
The Docker/noVNC approach for running Dwarf Fortress remotely demonstrates low-latency streaming of graphics-intensive applications. Similar technology is used for cloud-based AI inference (e.g., GPU-powered Jupyter notebooks, game AI). Why it matters: As AI models grow larger, running them locally on consumer hardware becomes impractical; remote streaming with client-side rendering (via WebGPU) will bridge the gap. Implication: Expect growth in services that stream AI model outputs (e.g., real-time video generation) directly to browsers without local compute.
Analysis generated by deepseek-reasoner