HackerNews Digest

The article argues that AI coding tools should be used for deliberate, high‑quality development rather than rapid, low‑quality output. It notes that current LLMs (e.g., Anthropic Claude, OpenAI models, Codex, Cursor Bugbot) are effective at identifying bugs across a codebase, though prioritizing and validating findings remains challenging. The author describes a workflow that runs multiple agents on a pull request, aggregates bug reports categorized as critical, high, medium, or low, and then manually verifies false positives before producing a final report. Critical and high‑severity bugs are fixed iteratively; medium‑level issues are addressed only when worthwhile, and PRs with excessive critical bugs may be abandoned. This process often uncovers pre‑existing defects, leading to additional unit‑test writing and refactoring, which can slow velocity but improves overall code health and developer understanding. The author encourages developers to adopt a slower, methodical “vibe coding” approach, using AI agents for comprehensive review, documentation, and architectural insight.

A study published in the APA’s *Journal of Experimental Psychology: Learning, Memory and Cognition* examined whether walking boosts creative thinking. Researchers Marily Oppezzo (Santa Clara University) and Daniel L. Schwartz (Stanford) tested 176 participants, primarily college students, using standard creativity tasks (alternate uses, analogy generation) and a convergent‑thinking word‑association task. Across multiple experiments, participants who walked on a treadmill or outdoors produced significantly more novel and appropriate responses than when seated or when seated in a wheelchair. Reported gains ranged from 81 % to 100 % higher novelty rates; one experiment showed a two‑fold increase in novel ideas while walking. Walkers performed slightly worse on the single‑answer word‑association task. A residual effect persisted: participants who walked before a second seated test generated more novel ideas than those who sat for both tests. The findings suggest that brief, moderate walking can temporarily enhance free‑flowing thought without requiring intense exercise. Funding came from the Knut and Alice Wallenberg Foundation, the Gordon and Betty Moore Foundation, and a Stanford Graduate School of Education dissertation grant.

Shamir’s Secret Sharing (1979) splits a secret into n shares such that any k shares can reconstruct the secret while any fewer reveal no information. The method treats the secret as the constant term of a degree‑(k‑1) polynomial over a finite field; random coefficients generate the remaining terms. Each participant receives a point (x, y) on the polynomial. With a single point the polynomial is under‑determined, so the secret could be any value. With k distinct points the polynomial is uniquely defined, allowing evaluation at x = 0 to recover the secret. A 2‑of‑n scheme uses a line (degree 1), a 3‑of‑n scheme a parabola (degree 2), and higher thresholds correspond to higher‑degree curves. The scheme guarantees information‑theoretic security: missing shares leave the secret uniformly distributed. Implementations replace geometric intuition with finite‑field arithmetic, enabling practical applications such as Ente’s Legacy Kit, where recovered shares are used locally in a larger, revocable recovery flow.

Norway’s National Library is building a sovereign Norwegian‑language large language model (LLM) to preserve the nation’s cultural heritage. The project, mandated by the Ministry of Culture, leverages the library’s digitized collection—20 PB of unique material stored in a 3‑2‑1 preservation system (≈60 PB total). Data preparation runs on an on‑premises AI pipeline that includes an Nvidia DGX H200, a 384‑core CPU cluster and multiple Huawei OceanStor Dorado all‑flash arrays providing 2 PB of low‑latency storage for ingestion, cleaning, deduplication and formatting. Processed data are then transferred to the national Sigma2 Olivia supercomputer (HPE Cray EX), which hosts 448 GPUs, 64 512 CPU cores and a 5.3 PB CrayClusterStor E1000 system for training. Key challenges identified are: moving petabyte‑scale data from a high‑durability archive to a high‑throughput AI pipeline; lack of standard evaluation tools for a bilingual, dialect‑rich Norwegian LLM; and governance/orchestration across the preservation archive, AI environment, and supercomputer. The initiative highlights Huawei flash storage’s significant role in European AI infrastructure.

None

The rollout of the new mitigation for Exit IP VPN servers includes the following nodes, each identified by location and internal code: - au-mel-wg-402 (Australia, Melbourne) - au-syd-wg-001 (Australia, Sydney) - ca-mtr-wg-302 (Canada, Montreal) - de-fra-wg-103 (Germany, Frankfurt) - fi-hel-wg-201 (Finland, Helsinki) - fr-par-wg-101 (France, Paris) - ie-dub-wg-101 (Ireland, Dublin) - no-osl-wg-101 (Norway, Oslo) - se-sto-wg-208 (Sweden, Stockholm) - us-dal-wg-701 (United States, Dallas) - us-lax-wg-002 (United States, Los Angeles) - us-nyc-wg-601 (United States, New York City) - us-slc-wg-303 (United States, Salt Lake City) These servers now have the mitigation applied, indicating an operational update across multiple geographic regions.

- The page lists a chronology of optimal “square‑in‑square’’ packings, each entry giving the side length \(s\) of the smallest containing square, the algebraic equation whose root yields \(s\), and the discoverer/date. - Early results (late 1970s–1980) include exact forms such as \(s=\tfrac72+\tfrac12\sqrt7\) (Bidwell, 1998) and \(s=3+\tfrac{4}{3}\sqrt2\) (Wainwright, 1979). - Polynomial roots of degree 5, 8, 18, 42, 82 etc. appear, e.g. \(9s^{5}-171s^{4}+999s^{3}-1959s^{2}+1636s+166=0\) (Schadt, 2026) and a degree‑42 equation (Schadt, 2025). - Since 2000 many entries were obtained by computer‑assisted methods: simulated‑annealing programs (Schadt, 2025‑26; Morandi, 2010), custom optimization scripts (Ellsworth, 2025‑26), and earlier programs by Cantrell and DeVincentis. - Several rational‑length packings are highlighted, notably \(s=7+\tfrac{4}{7}\) (Schadt, 2025) derived from the \(\{3,4,5\}\) Pythagorean triple, and the first doubly semi‑primitive packing. - The list shows continual improvement of bounds for square‑in‑square packings, with each new record either refining an algebraic value or achieving a lower‑degree polynomial representation.

The post recounts a FOO Camp session co‑led by the author and Matt Webb that applied Venkatesh Rao’s “Manufactured Normalcy Field” (MNF) to design. Rao’s MNF describes how people adopt new technology by minimally adjusting mental models, using familiar metaphors (e.g., smartphones as phones, the web as documents) and design choices that mask strangeness (e.g., commercial aviation’s controlled motion). The session split a whiteboard into three zones: “Things That Feel Weird” (technologies needing minimal normalizing), “Things That Feel Normal” (mundane items to be defamiliarized), and “Things We Use to Feel About Things” (cultural techniques for normalizing or denormalizing). Participants listed examples such as smile‑detecting chips, self‑driving cars, refrigerators, and crowdsourcing, and identified tactics like personification, skeuomorphs, gamification, and pathologizing. They then generated concepts to “weird” normal items (e.g., advertising in cemeteries, “Fridge as Narnia”) and to “normalize” weird ones (e.g., shared plane boarding controls). Ze Frank later featured the exercise, encouraging viewers to describe ordinary objects in ways that reveal their inherent oddness, linking the practice to Object‑Oriented Ontology’s flat‑ontology perspective.

None

Toshifumi Suzuki (b. 1 Dec 1932, Nagano) rose from a publishing‑sales background to lead Ito‑Yokado Group and its subsidiary Seven‑Eleven Japan as chairman and CEO. After joining Ito‑Yokado in 1963, he introduced franchising to Japanese retail, founding Seven‑Eleven Japan in 1973 and expanding it to over 10,000 stores by 2003, many operating 24 hours. Suzuki licensed the U.S. 7‑Eleven brand, imported its management systems, and deployed advanced point‑of‑sale and integrated data networks that linked sales, inventory, weather and supply‑chain information, enabling just‑in‑time deliveries and eliminating unsold‑goods returns. In 1991 he orchestrated a $430 million investment giving Ito‑Yokado 70 % of Southland Corp., then restructured the U.S. chain with the Japanese information system, restoring profitability by 1994. His management emphasized a flat hierarchy, regular cross‑level meetings, third‑party logistics, and aggressive use of IT. Beyond retail, Suzuki launched IY Bank, e‑commerce initiatives (7‑dream.com), and served in senior roles at Keidanren and environmental and ethics commissions. By 2003 the group generated over $28 billion in sales.