Published on May 13, 2026 at 18:01 CEST (UTC+2)
Setting up a free *.city.state.us locality domain (102 points by speckx)
Setting up a free *.city.state.us locality domain
This article explains how US residents can obtain a free locality domain (e.g., name.city.state.us) by registering through delegated managers. It provides step-by-step guidance on choosing a locality, setting up nameservers via Amazon Lightsail, and submitting the required form. The piece also covers the history of locality domains since 1992 and the eligibility requirements for US citizens or organizations.
Why I'm leaving GitHub for Forgejo (294 points by jorijn)
Why I'm leaving GitHub for Forgejo
The author describes moving their code from GitHub to a self-hosted Forgejo instance, driven by concerns over ownership and digital autonomy rather than outages. They highlight the Dutch government’s similar decision to launch code.overheid.nl on Forgejo for legal compliance and openness. The post details the migration process, including archiving public repos and pointing them to the self-hosted alternative.
I Moved My Digital Stack to Europe (565 points by monokai_nl)
I Moved My Digital Stack to Europe
This article chronicles the author’s migration from US-based cloud services to European alternatives, motivated by digital sovereignty and unpredictability of US jurisdiction. They replace Google Analytics with self-hosted Matomo and swap other SaaS tools. The narrative emphasizes choosing infrastructure based on values and control, not just convenience.
Reverting the incremental GC in Python 3.14 and 3.15 (82 points by curiousgal)
Reverting the incremental GC in Python 3.14 and 3.15
Python’s core development team announces the reversion of the incremental garbage collector shipped in 3.14 due to significant memory pressure in production. They will revert to the generational GC from 3.13 for both 3.14 and 3.15, with plans to reintroduce the feature via a proper PEP for Python 3.16. The decision was made to ensure stability, and patch releases will be expedited.
Kickstarter Is Forced to Ban Adult Content by Payment Processors (66 points by stalfosknight)
Kickstarter Is Forced to Ban Adult Content by Payment Processors
Kickstarter updated its mature content guidelines to prohibit a wide range of NSFW material, including implied nudity and specific terms like “MILF/DILF.” Reports indicate the change is driven by pressure from payment processor Stripe, which is partially owned by controversial figures. The article notes the impact on creators and the broader trend of platforms restricting content due to payment processor policies.
Preserving Fisher-Price Pixter (139 points by dmitrygr)
Preserving Fisher-Price Pixter
This technical deep dive details the complete reverse engineering, documentation, and emulation of Fisher-Price Pixter devices and their game cartridges. The author explains the challenges of dumping ROMs, emulating custom hardware, and preserving the audio and touch-screen features. The project covers multiple Pixter models, including Classic, Color, and Multimedia, and provides a file format for future preservation.
Dutch suicide prevention website shares data with tech companies without consent (190 points by giuliomagnifico)
Dutch suicide prevention website shares data with tech companies without consent
An ethical hacker discovered that the Dutch suicide prevention hotline 113 was sharing visitor data (location, browser, device, referral URL) with Google and other third parties without explicit consent. After being confronted, the foundation temporarily suspended all measurement tools. The story highlights severe GDPR violations and the sensitivity of data from vulnerable users.
An idiot's guide to lead optimisation for proteins (64 points by magni121)
An idiot's guide to lead optimisation for proteins
Written for beginners, this guide explains how machine learning is used in drug design for protein lead optimization. It follows the Cradle-1 pipeline, covering what proteins are and how ML models help improve candidate molecules. The author shares insights from experts, aiming to demystify the process for those new to computational biology.
New stainless steel can survive conditions for hydrogen production in seawater (204 points by HardwareLust)
New stainless steel can survive conditions for hydrogen production in seawater
Researchers at the University of Hong Kong developed a “super steel” (SS-H2) that resists corrosion in harsh electrolyzer environments, enabling green hydrogen production from seawater. The material uses a novel dual-passivation mechanism and could replace costly titanium parts. This breakthrough could lower costs for large-scale clean energy hydrogen systems.
Nailing jelly to a wall: is it possible? (2005) (26 points by microsoftedging)
Nailing jelly to a wall: is it possible? (2005)
This humorous experiment tests the proverb “it's like nailing jelly to a wall” by attempting to literally nail jelly to a wooden plank. The author describes the materials, preparation, and the failed (and messy) results. The piece is a lighthearted exploration of a seemingly impossible task, highlighting the difference between figurative language and physical reality.
Data sovereignty drives AI infrastructure choices
Multiple articles (leaving GitHub, moving stack to Europe) reveal a growing trend: developers and organizations are prioritizing control over their data and code over convenience. For AI/ML, this means training data, model weights, and inference pipelines may increasingly be hosted on sovereign or self-managed infrastructure. Implication: AI/ML practitioners should evaluate vendor lock-in and legal jurisdiction risks, especially when handling sensitive user data or complying with regulations like GDPR.
Ethical data handling is under intense scrutiny
The Dutch suicide prevention website incident shows how even well-intentioned services can violate privacy, with third-party trackers leaking highly sensitive behavioral data. For AI/ML systems that rely on web data for training or personalization, this underscores the need for rigorous consent mechanisms and data minimization. Implication: AI developers must audit data pipelines for compliance and avoid using data from sources with unclear consent — the reputational and legal fallout can be severe.
Performance vs. stability trade-offs in core AI/ML infrastructure
Python’s reversion of the incremental garbage collector highlights that even well-intentioned performance improvements can cause production issues (here, memory pressure). Many AI/ML frameworks (PyTorch, TensorFlow) depend on Python’s runtime. Implication: When adopting new runtime features (e.g., new GC, JIT compilers), teams should conduct extensive stress testing under realistic AI workloads before enabling them by default.
Open-source tooling gains traction for autonomy and compliance
The move to Forgejo (over GitHub) and self-hosted Matomo (over Google Analytics) reflects a broader shift toward open-source, self-hosted alternatives. For AI/ML, this aligns with the rise of open-source models (Llama, Mistral) and tools (MLflow, Kubeflow). Implication: AI/ML teams should consider self-hosted or community-driven platforms for version control, experiment tracking, and model serving to avoid platform risks and maintain full ownership.
Material science AI fuels experimental breakthroughs
The “super steel” discovery, while not AI-driven in the article, directly ties to the growing use of machine learning for materials design (e.g., predicting corrosion resistance, optimizing alloy compositions). The article mentions “cannot be explained” conventional theories, suggesting AI could help uncover hidden mechanisms. Implication: ML models trained on corrosion data or electrochemical properties can accelerate the discovery of new materials for clean energy, requiring interdisciplinary collaboration between domain scientists and AI researchers.
AI/ML for drug discovery requires accessible educational resources
The “idiot’s guide” to protein lead optimization illustrates the demand for clear, beginner-friendly explanations of how ML pipelines work in biology. As AI methods (diffusion models, protein language models) become central to drug design, the field faces a talent gap. Implication: Creating and sharing open educational content (like this guide) will help onboard more researchers and developers, democratizing access to cutting-edge AI-driven drug discovery.
Payment processor policies are reshaping content moderation in AI-dependent platforms
Kickstarter’s forced ban on adult content due to Stripe’s rules signals how financial infrastructure can indirectly control what content is allowed. For AI/ML, this affects platforms that generate or host AI-created content (e.g., NSFW image generators). Implication: AI startups relying on third-party payment processors should factor content restrictions into their business models, and consider decentralized payment alternatives or self-hosted crowdfunding to preserve flexibility.
Analysis generated by deepseek-reasoner