Adam Skye Jones

From linkAnkiDroid's maintainers say Google Play has rejected app updates over the project's Open Collective donation link, even after the team supplied documentation showing its fiscal host is tax-exempt. The issue explains that Google appears to be conflating a donor's tax deductibility with an organization's tax-exempt status, leaving the volunteer-run project stuck between removing its only visible funding link and risking delisting outside India and Russia on September 11. The write-up is useful because it is not a vague policy complaint: it lays out the exact timeline, the quoted policy language, and the narrow place where platform bureaucracy can destabilize routine open-source maintenance.

From commentsThe HN discussion centered on the ambiguity of Google's wording and the larger fragility of app-store distribution. Commenters kept returning to the same point: if the policy says tax-exempt donations are allowed, then AnkiDroid's host status should matter more than whether donors get a deduction, and Google's support responses read like multiple incompatible interpretations stitched together. The thread also widened into a platform critique, with people arguing that Android's nominal openness still leaves important volunteer software dependent on one repository owner's inconsistent enforcement.

From linkMithil Vakde frames ARC as a sample-efficiency problem and shows a small transformer pipeline aimed at learning puzzle structure with very little data instead of leaning on giant pretrained models or synthetic corpora. The post's pitch is not just the 44 percent score on ARC-AGI-1, but the fact that the experiment was cheap and fast enough to encourage repeated iteration, which the author treats as a path toward more accessible and scientifically legible AI progress. The result reads like a case for smaller, more controlled research loops: ARC remains unsaturated, humans still find it easy, and a focused model can sometimes outperform the sprawling systems that dominate current discourse.

From commentsThe comments were impressed by the creativity of the result but careful about the headline claim. Some readers were excited to see non-LLM work and liked the emphasis on puzzle-specific abstraction, while others argued that saying it “beats many LLMs” blurs an important distinction because ARC is neither a language task nor a measure of general-purpose assistant quality. The thread settled into a productive tension between admiration for the research direction and skepticism about benchmark framing, with several people treating that tension itself as evidence that efficient, interpretable experiments deserve more attention.

From linkMonica's founder says the open-source personal CRM is being rebuilt from scratch because nearly a decade of accumulated product decisions no longer matches the tool he would design now. Rather than presenting the rewrite as a purely technical modernization, the post goes back to the original motivation of remembering what matters about people and asks foundational modeling questions about relationships, reminders, activities, and what a “person” should even mean in the system. That makes the announcement more substantial than a framework migration: it is a product reset driven by a changed understanding of the domain as much as by the weight of old code.

From commentsThe HN thread immediately became a rewrite debate. Some commenters argued that LLMs and modern tooling have meaningfully lowered the cost of a clean restart, especially for a product of Monica's scale, while others pushed back that rewrites are still expensive unless the test suite is strong enough to keep regressions from leaking everywhere. A second line of discussion focused on the “personal CRM” idea itself, with people intrigued by the use case but uneasy about the ethics or social tone of systematizing relationships, which is exactly the kind of product ambiguity the post seems to be revisiting.

From linkFastpotify positions itself as a native, MIT-licensed Spotify client built in Rust with librespot, aimed at people who want the streaming service without the weight of a browser-engine desktop shell. The project foregrounds fast startup, lower memory use, local gapless playback, Spotify Connect controls, desktop media integration, and extra touches like a Winamp-style mini player and MilkDrop-compatible visualization. In practice the page is selling a specific philosophy as much as a feature list: music software should feel like a responsive desktop app again, not a sluggish web product that happens to play audio.

From commentsThe HN comments turned into a long-running indictment of the official Spotify client. People compared memory footprints, recounted years of keyboard-shortcut regressions and playback bugs, and treated Fastpotify as proof that the service does not inherently require a bloated interface layer to work well. The nostalgia in the thread was not just for old skins or Winamp references; it was for a time when desktop media players were expected to be fast, focused, and trustworthy, and many commenters saw this project as a small corrective to that lost baseline.

From linkMozilla is adding an optional built-in ad blocker to Firefox on iOS that uses Apple's WebKit content-blocker APIs plus EasyList to suppress many third-party ads and related trackers before they load. The post is careful about constraints: the feature is off by default, it will not block every ad, and it exists precisely because iOS does not let Firefox offer the same extension ecosystem it has on desktop and Android. The announcement matters less as a dramatic technical leap than as a pragmatic concession to platform rules, showing Mozilla embedding a privacy feature directly into the browser where user-installed tooling is structurally limited.

From commentsThe comments were supportive of the direction but immediately ran into rollout frustration and platform comparison. Several people said the feature was being described as launched even though it still had not appeared for them, while others compared it with Safari, Brave, Orion, and iOS content-blocker apps to argue that Mozilla is catching up rather than opening entirely new ground. Even so, the thread's underlying mood was that any built-in ad relief on iOS matters, because the mobile web has become noisy enough that users will tolerate only so much purity from browsers that cannot practically filter it.

From linkThe Restroom Archive is a catalog of public and semi-public restrooms documented with a level of architectural care usually reserved for galleries, menus, or transit maps. Each entry logs place, date, spatial details, materials, accessibility hints, and small observations about maintenance, light, or decoration, which turns an overlooked utilitarian space into a design object and an urban record. What makes the site stick is its seriousness: it does not present the bathroom as a joke prop, but as a recurring civic experience worthy of photography, notation, and comparative taste.

From commentsThe HN thread oscillated between earnest appreciation and bathroom-nerd comedy, but the appreciation carried it. People swapped stories about hunting for the best hotel or museum restroom while traveling, suggested integrations with mapping tools, and used the project as an excuse to complain about splashy sinks, hostile doors, and other universal restroom design failures. Beneath the jokes was a real consensus that the archive works because it notices something most people rely on constantly and almost never document well, which made the obsessive specificity feel charming rather than absurd.