Today’s Hacker News felt less interested in novelty than in control. The liveliest threads were about jurisdiction, ownership, and the quiet systems work that makes tools feel dependable.

Reflections

The day kept circling back to sovereignty, but in a practical register rather than an ideological one. People were talking about where their code lives, who owns the surrounding platform, and how much operational risk is hiding inside convenient defaults. Even the more forward-looking stories, from tiny tool-calling models to Linux compatibility work, had an infrastructural mood: narrower systems, clearer boundaries, fewer magic claims. The sharpest lesson came from the security thread, which was really about the cost of neglecting basic controls.

Themes

• Digital sovereignty is no longer just a policy term; it now shows up in choices about hosting, code forges, and cloud dependencies.
• Smaller, tighter systems looked unusually persuasive today, whether that meant a 26M tool-use model or a rollback to a known garbage collector.
• Old infrastructure still matters: delegated locality domains, editor conventions, and kernel primitives all shape modern software more than they advertise.
• Operational discipline kept beating abstraction, especially in the threads about access revocation, benchmarking, and production regressions.

I moved my digital stack to Europe (https://monokai.com/articles/how-i-moved-my-digital-stack-to-europe/)

Summary: The author describes moving major parts of a personal and business software stack to European providers, from analytics and hosting to other core services, in search of clearer jurisdictional guarantees and less dependence on U.S. platforms. The piece is not framed as purity politics so much as a practical exercise in reducing concentration risk while keeping a few unavoidable American services in place.

Discussion:

• Several readers liked the diversification instinct but warned against treating Europe as a sanctuary, noting that legal and political exposure does not disappear just by changing region.
• Cloudflare became the main compromise case: some saw it as a reasonable edge-layer exception, while others argued it undercuts the logic of the migration.
• A few commenters shared their own moves into European infrastructure and said the work was inconvenient but clarifying once the stack boundaries became explicit.

---

Show HN: Needle: We Distilled Gemini Tool Calling into a 26M Model (https://github.com/cactus-compute/needle)

Summary: Cactus open-sourced Needle, a 26M-parameter model built specifically for function and tool calling. The core claim is that tool use is often a narrower problem than general chat or reasoning, closer to selecting the right schema and assembling arguments, which makes a small, fast, local model plausible on ordinary consumer hardware.

Discussion:

• Readers asked how well the model distinguishes between similar tools and whether training across a limited set of categories transfers to messier real workloads.
• Others immediately saw uses in voice assistants and command-line wrappers, where natural-language argument parsing matters more than broad conversational ability.
• There was also skepticism around distillation itself, including whether providers can detect and degrade attempts to train smaller student models from frontier systems.

---

Leaving GitHub for Forgejo (https://jorijn.com/en/blog/leaving-github-for-forgejo/)

Summary: This post argues for moving a canonical git host off GitHub and onto a self-hosted Forgejo instance, not because GitHub is unreliable, but because ownership of the publication surface matters. The author connects that personal decision to a larger European interest in software infrastructure that is both fully open source and institutionally owned.

Discussion:

• The most common response was that federation is the real missing feature, because self-hosting alone still leaves collaboration fragmented.
• Many commenters used the thread to restate an older point: git was designed to be decentralized, but the social and review layers collapsed back into centralized services.
• Others focused on what is lost when leaving GitHub, especially the social graph and inbound contribution path, and pointed to bridge tools that try to soften that break.

---

Setting up a free *.city.state.us locality domain (2025) (https://fredchan.org/blog/locality-domains-guide/)

Summary: The guide explains how to register a free locality domain under the little-known city.state.us namespace, walking through delegated administrators, nameserver setup, and the paperwork that still powers this pocket of American internet governance. It doubles as a reminder that some of the web’s naming systems are older, stranger, and more civic than most people remember.

Discussion:

• Readers noted that some delegated locality registrars now have limited online flows, though the experience still seems inconsistent and fragile.
• The lack of WHOIS privacy for .us domains came up repeatedly as a serious downside, especially for people drawn to the namespace for personal projects.
• Others enjoyed the institutional archaeology of it all, including the rediscovery of familiar school and locality subdomains from earlier eras of the web.

---

Linux gaming is faster because Windows APIs are becoming Linux kernel features (https://www.xda-developers.com/linux-gaming-is-getting-faster-because-windows-apis-are-becoming-linux-kernel-features/)

Summary: The article argues that Linux gaming improvements increasingly come from kernel-level work, such as implementing Windows synchronization behavior more directly, instead of relying only on user-space translation in Wine or Proton. The result is a compatibility story that looks less like emulation heroics and more like ordinary systems engineering, even if it still falls short for some classes of games.

Discussion:

• A recurring demand was for harder evidence, especially frametime and latency comparisons against the same hardware on Windows rather than anecdotes about average FPS.
• Plenty of readers still reported real progress in everyday use, particularly for older titles and for gaming on hardware that already spends much of its time inside Steam and Proton.
• The biggest unresolved objection was anti-cheat, which remains a structural blocker for many competitive multiplayer games regardless of performance gains elsewhere.

---

Twin brothers wipe 96 government databases minutes after being fired (https://arstechnica.com/tech-policy/2026/05/drop-database-what-not-to-do-after-losing-an-it-job/)

Summary: Ars Technica covers a case in which two fired contractors allegedly used still-valid access to delete dozens of government databases immediately after termination. The real subject is not the theatricality of DROP DATABASE, but the failure of ordinary controls: access revocation, credential handling, and separation of duties all seem to have gone badly wrong at the same time.

Discussion:

• Many readers were stunned that privileged access apparently remained live long enough after the firing to make the attack possible at all.
• The report that thousands of passwords were exfiltrated triggered a second wave of disbelief, with commenters asking what sort of system architecture could expose that much secret material.
• The thread turned into a blunt reminder that account revocation has to happen before or during termination, not as a cleanup task afterward.

---

A History of IDEs at Google (https://laurent.le-brun.eu/blog/a-history-of-ides-at-google)

Summary: Laurent Le Brun traces how Google balanced a tightly standardized monorepo with a surprisingly fragmented editor culture, then slowly pushed toward more shared internal IDEs and web-based tooling. The essay is less about any single product than about the uneasy boundary between local developer preference and the coordination demands of a very large codebase.

Discussion:

• Former Googlers added historical texture, especially around how tool choices once varied sharply by language and team even inside the same repository.
• Some readers treated the piece as background to the current wave of cloud IDEs and AI-assisted coding environments, where the editor becomes more of a managed surface.
• Others argued the classic IDE may itself give way to something more task-oriented, with agents and remote execution displacing the older editor-centric model.

---

Reverting the incremental GC in Python 3.14 and 3.15 (https://discuss.python.org/t/reverting-the-incremental-gc-in-python-3-14-and-3-15/107014)

Summary: Python core developers say the new incremental garbage collector in 3.14 caused enough production memory pressure that both 3.14 and 3.15 will revert to the older generational collector used in 3.13. It is an unusually visible rollback for a runtime change, but also a healthy one: the project is choosing predictable behavior over theoretical improvement after real deployments surfaced trouble.

Discussion:

• Some operators said they had already traced suspicious memory growth in upgraded services back to the new collector.
• Many readers approved of the fast reversal but questioned how such a foundational runtime change shipped without a more formal process or broader evaluation.
• The discussion also became a comparison point with other runtimes whose garbage-collection changes tend to arrive behind more benchmarking, process, or gradual rollout.