HackerNews Digest

The Internet Archive’s mission is “Universal Access to All Knowledge,” delivered through a free, ad‑free digital library that relies on its own servers, storage, power, cooling, and staff. Maintaining 210 petabytes of data requires ongoing infrastructure funding, with the organization averaging about $25 per recurring donor. For September, the Archive offers a 2:1 matching program: any new recurring donation of $25 or more is tripled (e.g., $25 becomes $75, $50 becomes $150, $100 becomes $300). The matching is intended to boost reliable, year‑round support for server operations, book digitization, and web archive accessibility. Donors are invited to join the Monthly Giving Circle to sustain these services for future generations.

A 1024‑byte C program implements a minimal Python‑like interpreter. The source is stored in a fixed 999‑character buffer, with a global integer table for single‑letter variables. Parsing uses a recursive‑descent style directly on the source; expressions are evaluated while parsed, without an intermediate bytecode or AST. Control flow relies on indentation: run_block reads lines until indentation decreases, and loops re‑parse their bodies by jumping back to the condition position. Functions are handled similarly, saving the caller’s position and jumping to the function definition stored in the source. The interpreter supports integer literals, single‑letter variables, assignment, arithmetic (+ - * % with precedence), single comparison per expression (< > <= >= ==), truthiness, if/else, while (with optional else), for x in range(y) (with optional else), zero‑argument function definitions and calls (including recursion), print of literals or integer expressions, and comments. Size reduction uses single‑letter identifiers, globals, implicit int types (C89), ASCII codes instead of character literals, ternary and comma operators, and bitwise logic. The readable version is ~4800 bytes; aggressive golfing yields the final 1024‑byte code shown.

Switzerland’s federal government is piloting the replacement of Microsoft 365 on 3,000 workstations (≈7 % of its staff) with the open‑source “openDesk” suite, aiming to finish migration by the end of 2027. The pilot follows a proof‑of‑concept (“PoC BOSS”) in which 172 employees evaluated the platform; document processing and email performed well, while large‑scale video conferencing showed limitations. During the pilot the new tools run alongside Microsoft 365. The Federal Chancellery allocated CHF 9 million for this rollout and, if successful, may extend it to all 54 000 federal computers.

Key drivers cited by Professor Matthias Stürmer are: (1) risk of foreign access to data under U.S. cloud law, (2) operational dependence on a single vendor, and (3) rising proprietary licensing costs. The Swiss military’s Cyber Command plans a full migration to openDesk by October 2026 for similar sovereignty reasons.

The initiative aligns with the 2024 EMBAG law, which mandates default open‑source publication of government software, and the 2025 digital‑sovereignty strategy that seeks to reduce reliance on external suppliers. Microsoft is concurrently investing CHF 325 million in AI and cloud infrastructure in Switzerland.

Anubis is a server‑side protection mechanism that presents a JavaScript‑based proof‑of‑work challenge, similar to Hashcash, to deter large‑scale automated scraping. When the page cannot load its JavaScript, it indicates the server is overloaded or the Anubis challenge failed, prompting the user to reload. The system is designed so the computational cost is negligible for individual visitors but becomes costly for mass scrapers, thereby reducing scraping incentives. Anubis also serves as a temporary measure while the site develops more sophisticated fingerprinting techniques (e.g., detecting headless browsers through font rendering) to exempt legitimate users from the challenge. The protection requires modern JavaScript support; extensions that block or modify JavaScript, such as JShelter, must be disabled for the site to function properly.

Engrim is a local‑first, project‑scoped SQLite engine that provides a cross‑model episodic memory layer for AI agents (Google Antigravity, Claude Code, Cursor/Windsurf). It decouples project state from any single vendor, allowing seamless model switches while preserving decisions, constraints, and architectural context. Memory entries are stored in ~/.engrim/memory.db and indexed with SQLite FTS5 (bm25) and static model2vec embeddings, enabling hybrid lexical‑vector retrieval via a reciprocal‑rank fusion router. The core architecture comprises adapters/hooks for each agent, an agent‑provenance engine (origin_agent field), a hybrid router, and storage components (curated memories, full‑text search, vectors, and a flight‑recorder log). A 105‑session case study on a 50 k‑line algorithmic‑trading codebase consolidated >153 k tokens into <1 k tokens (<1 % of the context window), eliminating context amnesia and achieving zero test regressions. CLI utilities support auto‑detection, setup, JSON‑RPC MCP server, and CRUD operations (engrim add/recall/context/etc.). All data remain offline, permission‑restricted (0600), git‑ignored, and the project is MIT‑licensed (2026).

Meta has agreed to a settlement of up to $18 billion after U.S. states alleged its platforms harm children with addictive designs. The deal imposes restrictions on teen accounts—daily two‑hour limits, nighttime bans, parental linkage, and stricter age verification—though implementation remains uncertain and broader content concerns persist. Author Joan Donovan argues that these measures are insufficient and that Meta’s leadership, particularly Mark Zuckerberg, should resign to prioritize safety over profit. She draws parallels to tobacco regulation, citing past testimonies linking Meta’s engagement algorithms to addictive practices. Donovan notes Meta’s history of litigation and harms, including privacy breaches, mental‑health impacts on youth, amplification of hate speech, misinformation, and “infodemic” effects. Whistleblowers such as Arturo Béjar and Frances Haugen have documented internal knowledge of these harms. The article calls on shareholders to demand leadership change, suggesting that Zuckerberg’s wealth insulates him from fines and that a new governance structure is needed to protect users.

• On 24 August 2026 X Corp issued cease‑and‑desist letters demanding the permanent removal of all Nitter instances and the project repository. The Nitter maintainers announced, after legal counsel, that the project will continue and further details will follow.
• Nitter is a free, open‑source, privacy‑focused front‑end for Twitter that operates without JavaScript or advertising. All client requests are proxied through a backend using Twitter’s unofficial API, preventing IP and fingerprint tracking. The service is lightweight (≈60 KB vs. 784 KB for twitter.com) and offers RSS feeds, theme support, responsive mobile design, and AGPL‑v3 licensing that prohibits proprietary forks.
• Current roadmap items include re‑enabling embeds, implementing an account system with timeline support, adding tweet/profile archiving, and providing a developer API.
• Donation channels are listed (Liberapay, Patreon, Ko‑fi, BTC, ETH, XMR, SOL, $Nitter token, ZEC) alongside CI badges for tests, Docker builds, and license status.
• Community resources: a Matrix channel, personal contact email ([email protected]), and a legal/DMCA contact ([email protected]).

Babylonian cooking collection presents two reconstructed recipes:

• Lamb stew with beets (serves 2 or 15 bite‑size portions): diced mutton, rendered sheep fat, salt, beer, water, onion, arugula, Persian shallots, cilantro, cumin, red beet, leek, garlic; garnish with dry coriander seed and a paste of cilantro and kurrat (wild leek). Procedure: sear lamb in fat, sauté aromatics, add beet and greens, deglaze with beer and water, simmer ~1 h, finish with leek‑garlic paste and garnish; serve with steamed bulgur or naan‑bread.

• “Unwinding” barley sourdough stew (serves 2 or 15 bite‑size portions): barley seeds, warm water, salt, kurrat or spring onion, cilantro, garlic, leeks, toasted sesame oil, additional water and salt; sourdough bread (bappiru) made from toasted barley flour, fermented 12 h, baked at 375 °F, then crumbled. Procedure: soak and toast barley, grind to flour, ferment dough, bake, cool, crush. For broth, sauté garlic‑leek paste in sesame oil, add water and salt, simmer ~1 h, add fresh leeks and cilantro near end, stir in crushed bread before serving.

Both recipes credit Gojko Barjamovic and Nawal Nashrallah and aim to recreate Mesopotamian culinary practices.