tl;dr: Open source maintainers have traditionally treated vulnerability reports as special obligations—owing responsiveness and credit to researchers in exchange for their scarce insight and confidentiality. But in 2026, LLMs have made vuln discovery cheap and abundant, shifting the bottleneck from finding bugs to triaging them, while also eroding the value of embargoes since attackers can run the same tools. The author argues maintainers should now focus on rapid triage, remediation, and prevention (possibly via LLM analysis in CI), reserving "special" treatment only for high-severity reports or trusted researchers.

tl;dr: In 1963, Jerry began doodling a map of an imaginary city during a boring job, set it aside in 1983, then resumed decades later — it now spans 4,000+ 8x10 panels arranged in a rough circle. The project is governed by an evolving deck of ~100 instruction cards that dictate what to paint, collage, or alter on each panel, along with rules about layers (paint bands, collage, "city squares," Void, Red Dimension, etc.) that progressively replace earlier work. The result is a rule-based, semi-randomized generative art system executed by hand over 60+ years.

tl;dr: Tony Krueger, a longtime Microsoft Word developer, passed away. He's credited with making spell-check non-blocking and introducing the now-ubiquitous red (and later green) squiggly underlines for misspellings and grammar errors. He also ported Chip's Challenge to Windows by reverse-engineering the MS-DOS version without source code.

tl;dr: FUTO has released an open swipe typing system for their offline Android keyboard, trained on a public dataset of 1M+ QWERTY English swipes (released under MIT). The architecture uses three small models—a layout-agnostic encoder, a per-language ContextLM, and a layout-specific decoder—totaling ~2.5M parameters, achieving a ~4% top-4 fail rate (under 1% excluding OOV). Models are under the FUTO Model License (requiring user-visible attribution), with an accompanying GPL C++ inference/beam-search library.

tl;dr: Crysta.ai 3D-printed the author's Gaussian splat of an insect by voxelating it into a translucent "crystal" block, with each voxel containing a mix of inks and varying transparency. The author trained the splat using spherical harmonics at level 0 in linear space to suit the non-view-dependent printing process, though results came out slightly dark with some splat artifacts and clumpy fur. They suggest future improvements like voxelization previews and MagicaVoxel import/export support.

tl;dr: Summary not available

tl;dr: Summary not available

tl;dr: Vitamin D RCTs have refuted the magical correlations (no 30% mortality reduction), but the author argues skeptics overcorrected: trials consistently show hazard ratios slightly below 1 for cancer and all-cause mortality, and detecting modest-but-meaningful effects would require sample sizes far larger than any trial conducted. Combined with evolutionary evidence (ancestral levels ~115 nmol/L, pale skin evolving despite folate costs) and biology (vitamin D receptors throughout the body), supplementing if you have low-ish levels is probably worthwhile—even a HR of 0.96 would beat the cost of a daily pill.

tl;dr: Meta has paused its Model Compatibility Initiative (MCI), a tool deployed to US employees in April that captured mouse movements, keystrokes, click locations, and screen content to train AI systems on human computer use. The pause follows an internal security incident where MCI-collected data was accessible to other Meta employees; an initial fix on June 18 failed to hold, prompting further lockdowns. Employees had been petitioning against the program over privacy concerns, and opt-outs were only added after protests.

tl;dr: A Google employee was fired two months ago after creating a Google Workspace CLI tool that went viral, topping Hacker News and gaining thousands of GitHub stars and users. The author believes the termination stemmed from internal fears about AI agents disrupting Workspace, noting the irony that Google announced its own official Workspace CLI at Google Cloud Next just two days before the firing.

tl;dr: A NYT investigation found that the shift toward larger SUVs and pickups—with hoods now averaging 3 feet tall and significantly bigger blind zones—has caused roughly 3,000 additional pedestrian deaths from 2016-2024, accounting for about 10% of the 75% surge in pedestrian fatalities since 2009. Automakers have pushed bigger vehicles because they generate nearly all industry profits, while a 2009 rollover safety rule inadvertently worsened visibility by thickening A-pillars. Regulators have largely ignored the issue, betting instead on automatic braking technology that tests show is unreliable.

tl;dr: Summary not available.

tl;dr: Developers are increasingly building "harness loops" around coding agents—outer loops that queue work, judge results, and re-prompt models until tasks complete without human intervention. The author argues this works well for ephemeral tasks like porting, benchmarking, and security scanning, but produces overly defensive, hard-to-understand code when used for long-lived codebases, creating systems treated more like organisms than deterministic machines. Opting out may not be viable due to competitive and security pressures, so the real question is how to preserve human judgment, code legibility, and engineering standards in a loop-driven future.

tl;dr: AI platforms like OpenAI and Anthropic have been massively subsidizing usage—analysis shows $200/month subscribers can burn $8,000-$14,000 in tokens—while OpenAI lost $38.5B on $13B revenue in 2025, spending 44% on sales and marketing. As companies shift to token-based pricing, enterprise customers are experiencing 7x cost spikes, with some firms finding AI more expensive than human workers. Servicing the industry's projected $3T debt would require displacing roughly 27% of US jobs, making the path to profitability implausible even as OpenAI and Anthropic head toward IPOs.

tl;dr: Mistral released OCR 4, a document extraction model that returns bounding boxes, typed block classifications (tables, equations, signatures, etc.), and per-word confidence scores across 170 languages, deployable in a single self-hosted container. It claims top scores on OlmOCRBench (85.20) and 72% win rates in human preference tests against competitors, though Mistral notes benchmark scoring artifacts inflate apparent errors on math and multi-column docs. Pricing is $4/1k pages via API ($2 batch), $5/1k for Document AI, available through Mistral Studio, AWS SageMaker, and Microsoft Foundry.

tl;dr: Summary not available

tl;dr: Baidu released Unlimited-OCR, a document parsing model that extends DeepSeek-OCR for one-shot long-horizon parsing of single images, multi-page documents, and PDFs up to 32k tokens. It supports two image configurations (gundam and base) and ships with both Hugging Face Transformers and SGLang inference paths, including a batch script with concurrent requests against an OpenAI-compatible API. The model is available on Hugging Face and ModelScope, with an accompanying arXiv paper.

tl;dr: Steve Braithwaite has driven his 23-foot Big Banana Car over 250,000 miles since building it in 2008, and says he's been pulled over hundreds of times—usually by cops who just want photos or an excuse to chat. The latest stop happened in Billings, Montana, over a license plate issue, but he wasn't ticketed. Braithwaite now plans a "World Needs More Whimsy Grand Tour" to drive the banana through Central America and eventually around the globe.

tl;dr: Madison Square Garden compiled a document tracking activists who publicly criticized its use of facial recognition technology, collecting their tweets and comments for internal distribution. The document was discovered in a 45GB cache of data stolen by hackers and leaked online, then reviewed by 404 Media. The revelation underscores MSG owner Jim Dolan's reputation for targeting critics, while simultaneously deploying biometric surveillance on patrons.

tl;dr: The European Parliament's economic committee approved draft rules for a digital euro, a central-bank-backed electronic wallet aimed at reducing the eurozone's reliance on U.S. payment networks like Visa and Mastercard amid strained transatlantic ties. The ECB plans a 12-month pilot starting in late 2026 ahead of a full launch in 2029, with final legislative approval potentially coming by year-end. The project, six years in development, has gained urgency under Trump's second term, though it still faces opposition from some lawmakers and concerns from banks over deposit outflows.