Published on May 19, 2026 at 06:00 CEST (UTC+2)
Peter Salus has died (56 points by speckx)
Peter Salus has died – This announcement on the TUHS mailing list reports the passing of Peter Salus on May 15, 2026. Salus was a noted Unix historian, best known for his book A Quarter Century of Unix, which is considered essential reading for students of Unix history. The post appears in a thread also noting the death of Peter Neumann, another computing pioneer.
Click (2016) (229 points by andrewzeno)
Click (2016) – Clickclickclick.click is a minimalist web-based game or experience that revolves entirely around the act of clicking. The site’s single-word content “Click” and its URL suggest an interactive, possibly addictive, click-counting or reaction-time challenge. It has garnered 229 points on Hacker News, likely for its clever simplicity and addictive design.
The last six months in LLMs in five minutes (62 points by yakkomajuri)
The last six months in LLMs in five minutes – Simon Willison’s lightning talk from PyCon US 2026 summarizes rapid LLM developments since November 2025, which he calls an “inflection point.” During that period, the “best” model changed hands five times among providers like Anthropic, OpenAI, and Google. He uses a creative test—generating an SVG of a pelican riding a bicycle—to illustrate differences in model capabilities.
Regex Chess: A 2-ply minimax chess engine in 84,688 regular expressions (45 points by surprisetalk)
Regex Chess: A 2-ply minimax chess engine in 84,688 regular expressions – Nicholas Carlini presents a chess engine built entirely from a sequence of regular expressions. The engine executes 84,688 regex replacements in order to evaluate and make a legal chess move, using a 2-ply minimax search. The project is a playful demonstration that even pure regex can be used for non-trivial computation.
Anthropic acquires Stainless (389 points by tomeraberbach)
Anthropic acquires Stainless – Anthropic announces the acquisition of Stainless, a company specializing in SDK and MCP server generation. Stainless has powered Anthropic’s official SDKs since early days and helps other companies create language-native APIs. The deal aims to improve Claude’s ability to connect to external data and tools, reflecting Anthropic’s shift from models to agent-based systems.
Anyone on the Internet Can Ring Your Doorbell (15 points by jrdres)
Anyone on the Internet Can Ring Your Doorbell – Security researcher ABGEO analyzes a cheap smart doorbell from Temu and discovers critical vulnerabilities. An attacker can steal the doorbell from its owner’s account, impersonate the device to show fake video, and extract the home WiFi password via a debug port. The findings highlight severe security gaps in low-cost IoT hardware.
Pope Leo XIV’s first encyclical Magnifica humanitas to be published May 25 (141 points by cucho)
Pope Leo XIV’s first encyclical Magnifica humanitas to be published May 25 – The Vatican announces Pope Leo XIV’s first encyclical, focused on preserving the human person in the age of artificial intelligence. The document will be presented on May 25, 2026, with speakers including Cardinal Fernández and Christopher Olah, co-founder of Anthropic. This marks a major institutional statement on AI ethics from the Catholic Church.
We stopped AI bot spam in our GitHub repo using Git's –author flag (435 points by ildari)
We stopped AI bot spam in our GitHub repo using Git’s –author flag – Archestra’s CTO describes how AI bots flooded their open-source repository with low-quality comments and pull requests, drowning out genuine contributors. They implemented a countermeasure using Git’s --author flag to filter and block bot accounts. The post warns that AI-generated spam is degrading the quality of open-source collaboration.
Hyperpolyglot Lisp: Common Lisp, Racket, Clojure, Emacs Lisp (141 points by veqq)
Hyperpolyglot Lisp: Common Lisp, Racket, Clojure, Emacs Lisp – This is a side-by-side reference sheet comparing syntax, grammar, and features of four major Lisp dialects. It covers variables, functions, macros, libraries, and more for Common Lisp, Racket, Clojure, and Emacs Lisp. The resource is useful for developers working across different Lisp environments.
The Quiet Renovation at Bitwarden (572 points by DaSHacka)
The Quiet Renovation at Bitwarden – A detailed critique of Bitwarden’s recent changes, including a price hike for Premium disguised in fake monthly increments and a quiet CEO transition. New CEO Michael Sullivan has a background in private equity and M&A, raising concerns that Bitwarden may be preparing for a sale or cost-cutting. The article argues these moves erode the trust the company built as an open-source underdog.
The AI model competition is accelerating rapidly, with leadership changing hands multiple times in months – Simon Willison’s talk highlights that between November 2025 and May 2026, the “best” LLM shifted five times among Claude, GPT, and Gemini. This hyper-competitive landscape means developers must stay agile, benchmarking models frequently and building abstractions (e.g., agent frameworks) that can swap backends easily. The implication: no single provider will dominate for long, and multi-model strategies become essential.
Agentic AI is driving infrastructure acquisitions – Anthropic’s purchase of Stainless is a clear signal that the frontier is moving from standalone models to agent systems that interact with external APIs and tools. By owning the SDK and MCP generation pipeline, Anthropic can ensure Claude agents can seamlessly connect to any service. This trend suggests other AI labs will invest in tooling, connectors, and platform integration, making “agent-ready” infrastructure a key competitive advantage.
AI ethics and governance are entering mainstream institutional frameworks – Pope Leo XIV’s encyclical on AI, featuring a leading AI researcher (Christopher Olah), demonstrates that religious and ethical bodies are actively shaping AI discourse. This is not just academic; it influences regulation, public opinion, and corporate social responsibility. Companies should anticipate more ethical guidelines and prepare to align with human-centered principles, especially in Europe and global religious contexts.
AI-generated spam is poisoning open-source collaboration – The Archestra post shows that bots are flooding GitHub issues and PRs with low-quality, AI-generated content, overwhelming maintainers. The solution of using Git’s --author flag to filter by known human contributors is a stopgap, but the problem is systemic. Expect platforms like GitHub to develop better AI-detection tools and reputation systems. For open-source maintainers, proactive moderation and stricter contribution gates (e.g., verified accounts) will become necessary.
Cheap IoT devices remain a massive security vulnerability, often with AI implications – The smart doorbell hack reveals that low-cost hardware can be trivially exploited, giving attackers access to home networks and cameras. As AI agents increasingly control smart home devices (via voice assistants or automation), insecure IoT becomes a vector for larger attacks. The trend: AI security teams must include hardware and firmware analysis, not just software, and consumers need better awareness of supply chain risks.
The economics of open-source are shifting under private equity influence – The Bitwarden analysis shows how leadership changes and price hikes signal a move from community-driven to profit-maximizing governance. As AI companies rely heavily on open-source libraries and models, this trend could affect availability, licensing, and trust. Developers should monitor the business models behind their dependencies and consider forking or self-hosting critical tools.
Creative, low-resource AI demos continue to push boundaries of what’s possible – The Regex Chess engine is a whimsical but technically impressive example: using 84,688 regexes to play chess. While not practical, it demonstrates that AI/ML creativity can emerge from constraint-based programming. For practitioners, it’s a reminder that innovative solutions often come from rethinking fundamental tools, and that “AI” doesn’t always mean neural networks—symbolic approaches still have value.
Analysis generated by deepseek-reasoner