Windows 3.1 tiled background .bmp archive
Summary
The page references a GitHub repository named **andreasjansson/win-3.1-backgrounds**. The repository hosts an archive of bitmap (.bmp) files that reproduce the tiled desktop backgrounds used in Microsoft Windows 3.1. The content appears to be organized for download or cloning via standard GitHub interfaces. The surrounding page layout includes generic UI elements such as “Saved searches” and navigation headers, but no additional technical documentation, usage instructions, or code excerpts are provided in the scraped excerpt. The repository’s primary purpose is to preserve and distribute the original Windows 3.1 background images in their original BMP format.
Read full article →
Community Discussion
Comments express nostalgia for early PC wallpapers and tiled backgrounds, recalling personal experiences with Windows 3.x and early Linux textures. Many share enjoyment of the visual style and appreciate archived collections, while noting that some images appear garish or modern compared to typical desktop art. Several mention technical aspects such as resource impact on slow machines and difficulty finding original sources. Overall sentiment is fondness for the era, with a desire for preservation and access to these retro graphics.
Epoch confirms GPT5.4 Pro solved a Frontier Math Open Problem for the first time
Summary
A Ramsey‑style hypergraph problem concerning the sequence \(H(n)\) – the maximal vertex count of a hypergraph with no isolated vertices and no partition of size > \(n\) – was solved. The solution was first generated by Kevin Barreto and Liam Price using GPT‑5.4 Pro, then verified by problem contributor Will Brian, who plans to publish the result and may include follow‑up work. Brian noted that the AI‑derived construction eliminates an inefficiency in the lower‑bound argument and aligns closely with the existing upper‑bound structure, yielding tight bounds for this Ramsey‑theoretic setting. Subsequent testing on the FrontierMath “Open Problems” scaffold showed that additional models (Opus 4.6 max, Gemini 3.1 Pro, and GPT‑5.4 xhigh) also solved the problem. The original task asked for improved lower‑bound constructions for \(H(n)\), with a warm‑up for known cases, a single challenge for an unsolved \(n\), and a full problem seeking a general algorithm for all \(n\).
Read full article →
Community Discussion
The discussion centers on curiosity about the term “scaffold,” with participants interpreting it as a possible unit‑test framework for proofs and requesting clarification of its purpose. The tone is inquisitive rather than evaluative, focusing on understanding the methodology used for testing models on FrontierMath. No strong agreement or disagreement is expressed beyond the desire for more detail.
Autoresearch on an old research idea
Community Discussion
The comments recognize that LLM‑driven autoresearch can automate bug fixing and hyper‑parameter tuning, often yielding modest performance gains at relatively low effort. Users note that the approach is effective for optimization but rarely produces novel architectural ideas, and its value depends heavily on the quality of evaluation metrics and available data. Cost of token usage and compute is seen as a concern, while some view it as a useful substitute for repetitive consulting advice. Overall sentiment is cautiously optimistic about its utility for routine engineering tasks but skeptical about broader research impact.
FCC Updates Covered List to Include Foreign-Made Consumer Routers
Community Discussion
The comments focus on the FCC’s new router restrictions, expressing skepticism that security issues stem from foreign manufacturing rather than poor industry practices and weak regulatory oversight. Many note the difficulty of defining “produced abroad” and question the practicality of a domestic‑made router market, while suggesting open‑source or DIY hardware as a more trustworthy alternative. Critics view the policy as a protectionist or surveillance‑oriented measure that could raise prices and limit consumer choice, and they seek clearer guidance on exemptions and compliance.
iPhone 17 Pro Demonstrated Running a 400B LLM
Summary
The page displays an error notice stating that an operation failed and invites the user to retry the action. No substantive textual content, data, or technical information is provided beyond this generic message. An accompanying image is referenced only by its alternative text, which consists of a warning emoji (⚠️). No further context, instructions, or details are included in the scraped material.
Read full article →
Community Discussion
The comments recognize the technical novelty of running a large‑scale, mixture‑of‑experts model on a mobile device, noting the impressive engineering behind the demonstration. However, many point out practical limitations such as high power draw, heat, limited token throughput, and modest real‑world use cases, questioning the immediate usefulness of a 400 B‑parameter model on a phone. Skepticism is expressed about quantization claims and whether such feats represent genuine progress or mainly hype, while others anticipate future improvements in hardware, memory, and model efficiency that could eventually make mobile inference more viable.
Pompeii's battle scars linked to an ancient 'machine gun'
Community Discussion
The comment presents factual information about the polybolos, describing it as an advanced ancient Greek repeating ballista invented by Dionysius of Alexandria, and noting its chain‑drive and gravity‑fed mechanism that allowed rapid successive firing. It emphasizes the device’s historical significance, often labeling it a “machine gun of antiquity.” The overall tone is neutral and informative, focusing on technical details without expressing personal opinion or strong sentiment.
The Resolv hack: How one compromised key printed $23M
Summary
The Resolv DeFi protocol was exploited on March 22 2026 when an attacker compromised the protocol’s AWS KMS environment and obtained the privileged private key used by the off‑chain SERVICE_ROLE to sign mint‑approval transactions. The minting design required only a valid signature, without any on‑chain cap or collateral‑to‑mint ratio check. Using the stolen key, the attacker deposited roughly $100‑200 k USDC and called the `completeSwap` function to mint about 80 million USR (≈ $25 M) far exceeding the collateral. The USR was then wrapped into wstUSR, swapped through stablecoins, and converted to ETH, resulting in an estimated 11,400 ETH (≈ $24 M) extracted. The sudden influx of unbacked USR caused the token’s peg to collapse by ~80 % before partially recovering. Resolv halted operations and began investigation. Post‑mortem analysis highlights that the vulnerability stemmed from reliance on off‑chain key management rather than smart‑contract logic, and that real‑time on‑chain monitoring (e.g., anomalous mint‑to‑deposit ratios) could have generated early alerts or automated pauses to limit loss.
Read full article →
Community Discussion
The comments focus on a recent hack in which an attacker compromised a platform’s AWS account, stole a signing key, and minted large amounts of tokens before the protocol was halted, emphasizing that the breach stemmed from traditional credential theft rather than a pure smart‑contract flaw. The discussion repeatedly questions the usefulness and stability of stablecoins, critiques the industry’s vulnerability and regulatory gaps, and speculates about an insider’s involvement, while expressing overall skepticism toward crypto’s security and practical value.
Show HN: Cq – Stack Overflow for AI coding agents
Summary
The post introduces “cq,” a proposed open‑source commons for AI coding agents that functions like a Stack Overflow for agents. It argues that LLMs trained on Stack Overflow data have begun to erode the community that supplied that knowledge, leading to repeated token‑heavy failures when agents work in isolation. cq aims to reduce waste by enabling agents to query a shared knowledge store before tackling unfamiliar APIs, CI/CD configurations, or frameworks, and to contribute verified solutions back to the pool. Trust is built through multi‑agent confirmation, confidence scoring, and reputation signals, addressing the reported drop in confidence (46% of developers distrust AI output). The early implementation includes plugins for Claude Code and OpenCode, an MCP server for local storage, a team API, a human‑in‑the‑loop UI, and containerized deployment. Mozilla AI is positioning cq as a standard‑setting, open‑source effort to keep AI tooling transparent, collaborative, and sustainable across organizations.
Read full article →
Community Discussion
The discussion acknowledges the concept’s promise while emphasizing significant security and trust challenges. Commenters highlight risks of malicious content injection, the difficulty of establishing robust, decentralized reputation systems, and the need for personalized trust metrics to resist manipulation. Suggestions include leveraging web‑of‑trust frameworks, EigenTrust, and delegated authentication for agents. Opinions range from cautious optimism about corporate‑level applications to skepticism about practical reliability, noting that thorough safety engineering and thoughtful schema design are essential before broader adoption.
Abusing Customizable Selects
Community Discussion
The comment expresses strong appreciation for the author’s creativity, describing each item as more engaging than multiple Google tools. It also reflects a perspective that industrial tools are designed to reduce novelty, implying that their purpose is to limit surprise. Overall, the sentiment is positive toward the creative work while offering a critical view of the conventional, utilitarian nature of industrial tools.
Ju Ci: The Art of Repairing Porcelain
Summary
Ju ci (锔瓷) is a Chinese porcelain‑repair craft originating in the Song dynasty (960‑1279), documented in 13th‑century paintings. Recognized by UNESCO as intangible cultural heritage, the technique restores broken ceramics by drilling precise holes and inserting handcrafted metal staples—copper, iron, gold, or silver—into the fractures. Artisans align the staples to both secure the pieces and preserve the object's visual integrity, allowing functional use while highlighting the repair. Ju ci shares a philosophical stance with Japanese kintsugi, emphasizing the aesthetic value of imperfections and treating repaired cracks as narrative elements that confer a renewed identity on the object. The method integrates meticulous metalwork with traditional ceramic conservation, reflecting both technical skill and cultural values of resilience.
Read full article →
Community Discussion
The comments express strong appreciation for the visual result and the inventive tension‑based repair approach, while noting practical challenges such as timing the resin set and achieving sufficient clamp pressure. Viewers report surprise at the technique differing from expectations, and many wonder about the cup’s usability after repair, questioning whether it could be reglazed or refired. References to traditional Kintsugi highlight interest in comparable methods, and overall the discussion balances admiration for the craftsmanship with curiosity about functional durability.