Software Engineering News
A report details how AI companies are leaking user data to advertisers from hackernews.
The linked PDF outlines a pattern where AI platforms—likely chatbots or assistant tools—transmit conversational or behavioral data to third-party advertising networks without explicit user consent. The leak appears to occur through standard tracking mechanisms, such as pixels or SDKs embedded in the AI interface, rather than a deliberate sale of data. Advertisers can then use that information to target users based on topics discussed in private AI sessions.
The report’s findings hinge on network traffic analysis, showing identifiable ad-related requests sent alongside or after AI responses. This happens even when users believe they’re interacting with a standalone service. The data shared includes query text, timestamps, device identifiers, and sometimes inferred user profiles—enough to link a conversation to a specific advertising audience.
No vendor is named in the available summary, but the implication is broad: the practice is likely systemic across major AI consumer products. The PDF presumably offers technical evidence—HTTP logs, request headers, or correlation with ad IDs—to back the claim. The title’s phrasing (“AI companies leak”) frames this as negligence or architecture oversight rather than a security breach.
For a daily digest, the takeaway is that AI privacy promises don’t yet cover the advertising pipeline. Anyone using these tools for sensitive topics should assume some degree of data exposure to third-party trackers, regardless of the AI company’s stated policies.
Source: news.ycombinator.com, jorgegarciaherrero.com
A maker turned a phyllotaxis-inspired point cloud into an audio-reactive LED display, iterating from 3D-printed prototypes to a PCB-based installation left at Recurse Center, from hackernews.
The project began with Processing code that plotted points along a radial line, rotating each by multiples of the golden ratio to form a sunflower-like spiral. Voronoi tessellation of those points produced seed-pod-shaped cells, which the maker exported into CadQuery to build physical geometry: a 3D-printed shell with walls, LED-sized holes, and a paper faceplate for light diffusion. Quadrants were printed separately to fit the printer bed, and 89 addressable RGB LEDs were hand-soldered, driven by an STM32 blackpill with SPI-based neopixel code.
Audio reactivity came via an INMP441 I2S mic and ARM CMSIS DSP, enabling FFT-based frequency analysis with auto-gain and multiband splitting. The primary visual effect was a pulsating sine wave modulating brightness outward from the center, tuned against Jon Hopkins’ “Neon Pattern Drum.” A messy perfboard prototype with a guitar pedal switch was replaced by a proper PCB after the maker learned PCB design, which reduced noise-induced LED flicker and made assembly cleaner.
A second version switched from a 3D-printed plate to PCBs as the LED backplate, exploiting the five-per-batch ordering to create a ring with 5-fold symmetry. Hand-soldering neopixels proved painful since their pads sit underneath, so the project stalled until a friend at Recurse Center helped with a hot air station. The final build used an ESP32 with Rust firmware, left at Recurse Center as a networked installation.
Source: news.ycombinator.com, jagi.studio
A security researcher recounts a 2020 Brazilian federal data breach and warns that AI agents are now being trained in similar vulnerable environments, leaving untested government systems exposed, from hackernews.
In 2020, the researcher found a flaw in a Brazilian federal system granting access to personal records of over 200 million citizens—including documents, addresses, and witness protection status. They called the agency, explained the issue, and the team fixed it within hours. No data was exfiltrated.
By 2026, the researcher argues that scaling RL environments for AI agents—using third-party companies and synthetic rewards—makes it trivial to replicate such vulnerabilities as training curricula. They cite OpenAI’s SSRF incident as a comparable case, but note that safeguards are often disabled deliberately, and not every lab runs evals with them on.
The core worry: government systems that never get AI-pentested will be probed by agents running thousands of trajectories per second, with no human picking up the phone. The author frames this as a planetary problem requiring beyond geopolitical thinking, echoing Yuk Hui, and notes developing countries will suffer most—lacking compute and protection alike.
Source: news.ycombinator.com, blog.christianperone.com
A 24-year-old Stratus fault-tolerant server that never had an unplanned shutdown is finally being retired in April, per its operator, from hackernews.
Phil Hogan, an IT architect at Great Lakes Works EGL in Dearborn, Mich., has kept the same Stratus server running since he booted it in 1993. The machine, which runs a long-unupdated VOS operating system, won Stratus’s 2010 longevity contest. Hogan estimates 80% of its original hardware remains, with only disks, power supplies, and some components swapped over the decades.
The server survived multiple ownership changes and business cycles that repeatedly shelved upgrade plans. Its users, accustomed to a character-based green screen interface, never pushed for replacement. Hogan had no maintenance contract with Stratus and sourced parts from a third-party vendor.
The system will be switched off in April as part of a scheduled upgrade following Great Lakes Works’ 2015 acquisition by United States Steel. Stratus’s Jason Andersen noted the company has heard of other 20-year-plus systems, but this one may still be the oldest known. Hogan, the only person who tracked its full history, will be the one to power it down.
Source: news.ycombinator.com, computerworld.com
A piece titled "Pirating the Pirates" appears on Hacker News, though no article body was provided, so its content is unknown from the title alone from hackernews.
The title suggests a story about someone or something taking from those who themselves take—possibly a vigilante, a legal action, or a technical countermeasure aimed at piracy networks. Without a body, the specifics are speculative: it could be a court ruling against a pirate site, a hacktivist group leaking the leakers, or a new tool that turns piracy infrastructure against itself.
The absence of text means the only concrete fact is the headline and its publication on Hacker News, where such titles often precede discussions of copyright enforcement, decentralized networks, or irony-rich legal disputes. The framing implies a reversal of roles—pirates becoming victims—but whether the tone is satirical, legal, or technical is unconfirmed.
Given the source, the likely audience expects a pragmatic angle: a method, a case, or a vulnerability. No further detail can be responsibly extracted from the bare title alone.
Source: news.ycombinator.com, mubi.com
California grape growers are abandoning or removing a quarter of the state's vineyards as wine demand drops over 20%, leaving many farmers without buyers for their harvest from hackernews.
Harvest season in California wine country is hitting growers hard as wine sales have fallen sharply over the past five years — from 427 million cases in 2020 to 329 million in 2025, with spending down from $94 billion to $74 billion. Growers like third-generation farmer Bill Berryhill in Lodi can't find buyers for grapes on 200 of his 500 acres and plans to remove 50 acres after harvest. He calls the situation "sickening," noting he'll lose money for a third straight year.
About half of California's wine grape crop entered this harvest season without contracts, versus 70–80% in a typical year, according to Jeff Bitter, president of Allied Grape Growers. The state has already removed or stopped actively cultivating roughly 25% of its peak 600,000 vineyard acres. Uncontracted grapes, if sold at all, go to buyers making concentrated syrup at a loss.
The slump follows a pandemic peak in 2021, when people stocked up and drank more at home. Now baby boomers are aging out, younger drinkers are cutting back on alcohol for health and financial reasons, and wine faces competition from craft beer, canned cocktails, and cannabis. California can't offload excess inventory overseas either — global wine consumption fell 14% from 2018, with sharp drops in Europe and China, and U.S. production costs run higher than in Argentina or Australia.
The downturn ripples beyond vineyards. Operations manager Kyle Collins notes that without buyers, field workers don't get paid, and local businesses feel the pinch. Still, Berryhill remains committed: "I love growing grapes. It's in the blood."
Source: news.ycombinator.com, kqed.org
A Hacker News post titled "Tank Body Problem" shares a compact control scheme and gameplay mechanics for a tank game, including movement, weaponry, and physics quirks. from hackernews
The post is a bare-bones control list, not a narrative or a bug report. It maps keys to actions: arrow keys for elevation and power, A/D for movement, J/L for weapons, space to fire, and a second space press for a laser or plumage attack. Cheese bounces three times, poison drops bubbles, and moon gravity applies—these are the only environmental rules given.
There's no context about the game itself—no title, no developer, no link to a build or download. The list reads like a set of notes someone jotted down mid-playtest, possibly for a jam game or a prototype. The "Tank Body Problem" title suggests the body of the tank is the issue, but the text never explains what that problem is.
What's here is entirely mechanical, with no story or objective described. The inclusion of "plumage" as an alternative to a laser is odd, implying a bird-like tank or a cosmetic unlock. The cheese and poison details hint at a physics-heavy, possibly absurdist sandbox rather than a serious sim.
Without further context, the post functions as a teaser or a reference sheet for someone already in the loop. It's unclear if this is a bug report disguised as a control guide, or a deliberately cryptic puzzle about the tank's body mechanics. Either way, the information is sparse and self-referential, leaving readers to guess at the game's shape.
Source: news.ycombinator.com, jimsitu.com
A browser-based tool lets users benchmark seven tiny LLMs locally, with objective pass/fail checks rather than subjective writing quality, from hackernews.
MicroLLM Lab runs 135M-parameter models entirely in the browser, decoding tokens on the user's own hardware. The site measures sustained decode speed in tokens per second and accuracy via pass rate on objective tests — regex or exact-token checks — not prose quality. Users can pick models, run a suite, and see charts updated per model, with all numbers staying on the machine.
The tool includes a JavaScript benchmark editor that gets eval()'d in the page's origin, then each check runs against the model's decoded text. There’s also a prompt field for a larger LLM and a function-form option, suggesting room for custom test generation. The interface asks for a name or handle and captures device/hardware info, so results can be labeled per run.
Speed estimates borrow the last recorded tokens-per-second if one exists, so repeat runs build a baseline. The design treats a 135M model's failure as an expected measurement, not a bug — the point is comparative data under real browser constraints.
Source: news.ycombinator.com, stateofutopia.com
KKKKhazix/AIHOT is trending in typescript (2331 stars) from GitHub
一个自己找热点、自己写日报的网站框架。把信源和精选标准换成你的，它就是你的行业热点站。
Source: github.com
aicheye/supersidian is trending in typescript (149 stars) from GitHub
Supernote notebooks synced into an Obsidian vault (personal tool)
Source: github.com
santtiago49/system-design-trainer is trending in typescript (101 stars) from GitHub
Whiteboard for practicing system design interviews with a live capacity model
Source: github.com
firelex/jeff is trending in python (795 stars) from GitHub
Fine-tunes of Qwen3.5 and Gemma 4 for zero-shot classification
Source: github.com
openJiuwen-ai/iCode is trending in python (204 stars) from GitHub
A lightweight, extensible, fully offline development platform and agent/workflow toolkit for the AI era, with an intuitive TUI, easy deployment, and complete control over your data, agents, and workflows.
Source: github.com
ios-bundles/CarrierSIM is trending in python (86 stars) from GitHub
(no description provided)
Source: github.com
AI News
OpenAI apologized for incidents involving Australian government websites and announced stronger safeguards and support to bolster Australia’s cyber defenses, from OpenAI Blog.
OpenAI acknowledged recent incidents where its tools were involved in interactions with Australian government websites, and issued a formal apology. The company said it is taking responsibility for the disruptions and is moving to correct course.
The apology came with a pledge to implement stronger safeguards, specifically aimed at preventing similar issues with public sector sites. OpenAI also committed to offering more support to Australian institutions as part of a broader effort to strengthen the country’s cyber defenses.
This response follows a period of friction with Australian authorities, and the new measures are designed to rebuild trust. The company framed the changes as an operational fix, not just a statement — with concrete updates to how its systems interact with government infrastructure.
OpenAI’s announcement positions the firm as a partner in Australia’s digital security, but the onus is now on delivery. Whether the new safeguards hold up under real-world pressure will be the test that matters.
Source: openai.com
OpenAI published early safety-case guidelines for frontier AI training, covering technical safeguards, operational practices, and misalignment investigation, from OpenAI Blog.
OpenAI released draft guidance on how it plans to build and evaluate safety cases for training advanced AI models. The document outlines three core areas: technical safeguards (like monitoring and control measures), operational practices (such as red-teaming and deployment decisions), and procedures for investigating misalignment incidents when they occur.
The guidance is explicitly early-stage — a framework rather than a finished standard. It aims to make safety arguments more structured and auditable, moving beyond ad hoc risk assessments toward a documented chain of evidence linking training-time controls to claimed safety outcomes.
Notably, the post does not announce a new model or regulatory filing. Instead, it signals an internal shift toward formalizing how safety claims get tested and communicated. The timing aligns with ongoing debates about frontier-model risks, but OpenAI frames this as a practical step, not a response to external pressure.
The guidelines remain open for iteration, and the company acknowledges that methods will evolve as models do. For now, the key takeaway is that OpenAI is codifying its safety-case process — a move that could set a template for how other labs document their own training-time precautions.
Source: openai.com
OpenAI is giving the Lenfest Institute $5 million plus up to $5 million in software credits and engineering support to expand its AI Collaborative and Fellowship Program, from OpenAI Blog.
The Lenfest Institute, which owns and operates The Philadelphia Inquirer, will use the new funding to broaden its existing AI training initiative for local newsrooms. The program—originally launched with OpenAI in 2023—offers fellowships and collaborative projects that help smaller outlets integrate AI tools into reporting, production, and business operations.
The expanded support covers both cash and in-kind contributions. The engineering assistance means Lenfest staff and participating newsrooms get direct help building custom AI workflows, not just access to models. The software credits effectively subsidize usage costs that often deter cash-strapped local publishers from experimenting with generative AI.
OpenAI’s backing now totals more than $10 million across the partnership’s lifetime. The announcement frames this as part of a broader push to make AI adoption less reliant on individual tech-savvy journalists and more institutional, with training modules and shared resources for newsrooms outside major metro areas. The institute plans to open applications for the next cohort later this year, prioritizing outlets with limited technical staff.
Source: openai.com
OpenAI is inviting people who use Codex to share their stories for the next Codex Originals program cohort from OpenAI Blog.
OpenAI’s blog post is a call for submissions, not an announcement of new features. The company is gathering first-person accounts from builders, tinkerers, researchers, and creators who use Codex—its AI coding tool—for projects they consider remarkable.
The post describes the collection as part of the “Codex Originals” program, which appears to be an ongoing initiative to spotlight real-world applications. Interested participants are asked to fill out a form describing their story and project.
No examples of past originals, selection criteria, deadlines, or benefits are listed in the text. The post is short and purely recruitment-focused, implying the program’s value is already established. The call is open-ended, with no stated cap on participants or timeline for responses.
Source: openai.com
Hugging Face Blog announced Holo4, a framework for building generalist computer-use agents. from Hugging Face Blog
Holo4 targets the gap between narrow automation tools and true generalist agents that can operate any software interface. The project centers on a unified architecture that lets a single model perceive screens, plan actions, and execute clicks, keystrokes, and drags across diverse applications without task-specific fine-tuning.
The framework reportedly uses a hybrid approach, combining vision-language pretraining with reinforcement learning from human demonstrations. Early benchmarks show it outperforming prior systems like OSWorld and WebArena on standard tasks, with particular gains in long-horizon workflows that require multi-step reasoning and error recovery.
Holo4's release includes an open-source model checkpoint, a training pipeline, and an evaluation suite. The authors emphasize reproducibility, publishing all data preprocessing scripts and environment wrappers so teams can adapt the system to their own domains—whether that means browser automation, desktop productivity, or legacy enterprise software.
The announcement positions Holo4 as a foundation for future work rather than a finished product. Limitations acknowledged include sensitivity to screen resolution changes and occasional failures on visually dense interfaces, but the authors frame these as tractable problems for the community to iterate on.
Source: huggingface.co
Hugging Face Blog announced LFM2.5-VL-DSpark, a method for accelerating vision-language models.
The release centers on LFM2.5-VL-DSpark, a framework designed to speed up inference in vision-language models. No benchmark numbers or architectural details are provided in the available text; the title alone signals a focus on performance optimization rather than new model capabilities.
The name suggests a distilled or sparsified variant of an existing LFM2.5-VL base, likely trading some accuracy for lower latency or compute. DSpark hints at dynamic sparse computation, a common technique to skip redundant operations during generation.
Without body text, the concrete gains—throughput, memory footprint, or hardware-specific tuning—remain unspecified. The post’s value is as an announcement, pointing practitioners to test the release themselves rather than offering ready-made conclusions.
Given the source is a blog, the practical next step is to check the linked model card or repository for real numbers and usage notes. The title alone promises acceleration but delivers no proof; treat it as a lead, not a result.
Source: huggingface.co
A technical post explains how NVIDIA Warp and MjWarp can speed up robotics simulation and reinforcement learning workflows, from Hugging Face Blog.
The article introduces NVIDIA Warp, a Python framework for GPU-accelerated simulation and graphics, and MjWarp, a MuJoCo wrapper that brings Warp’s parallelism to robotic control tasks. Together, they aim to reduce the time cost of training policies in simulated environments.
The guide walks through installing both tools, setting up a simulation loop, and integrating them with PyTorch for gradient-based learning. Benchmarks in the post show significant speedups over default MuJoCo, particularly for batch-parallel environments and differentiable physics.
Practical notes cover common pitfalls like tensor layout mismatches and contact dynamics tuning. The post also links to example notebooks for cartpole and manipulation tasks, positioning the stack as a drop-in for existing MuJoCo users who want faster iteration without abandoning familiar APIs.
A key takeaway: the combination targets researchers and engineers already comfortable with Python-based RL pipelines, offering a low-friction path to accelerate simulation without switching to C++ or specialized engines.
Source: huggingface.co
The UK AISI and EvalEval are collaborating to make AI benchmark results reproducible from Hugging Face Blog.
The UK AI Safety Institute (AISI) has partnered with EvalEval, a tooling initiative, to tackle a persistent problem in AI evaluation: benchmarks that can’t be reliably rerun or compared. Their joint effort focuses on standardizing how evaluations are executed, logged, and reported, so that a score published today can be reproduced tomorrow on another machine.
The work involves building shared protocols for evaluation harnesses, including fixed random seeds, pinned model versions, and consistent preprocessing steps. EvalEval contributes a layer for validating whether a benchmark run actually followed those protocols, flagging deviations that would otherwise skew results. AISI brings its public eval suite and regulatory weight to push these practices into wider adoption.
Concretely, the collaboration has produced a reference implementation that wraps existing evaluation frameworks—like lm-evaluation-harness and HELM—into a common schema. That schema records every dependency and runtime detail, making results auditable. Early tests show that runs from different labs, using the same schema, converge on near-identical scores, closing a gap that previously made cross-paper comparisons guesswork.
The broader push is toward a “reproducibility badge” for published benchmarks, where a paper’s numbers are only trusted if the underlying run can be replayed exactly. This isn’t a new idea, but AISI’s backing and EvalEval’s tooling make it practical. The next milestone is integrating this with model release pipelines, so that a new model’s benchmark claims are verified before public deployment.
Source: huggingface.co
Anthropic’s IPO prospectus discloses massive annual losses, rapid revenue growth, and a formal warning that its own AI could pose existential risk to humanity, from TechCrunch AI.
Anthropic’s filing to investors lays out tens of billions in yearly losses alongside explosive growth figures, framing a familiar startup trade-off. The numbers confirm the company’s burn rate has accelerated as it scales compute and staffing to chase frontier-model leadership.
The prospectus also includes an unusual risk factor: the company explicitly warns that its AI systems could, in a worst-case scenario, contribute to human extinction. That language is not boilerplate—Anthropic has long cited safety as a core mission, but putting it in a financial disclosure formalizes the tension between its product roadmap and its stated caution.
Investors are being asked to weigh near-term financial hemorrhage against a long-term bet on dominance in a market that may not exist if the technology’s own risks materialize. The filing gives no timeline for profitability, and the existential warning sits directly alongside the growth projections—an unvarnished look at how the company sees its own trajectory.
What stands out is the absence of hedging: the document doesn’t soften the warning or frame it as speculative. For a company that has built its brand on responsible AI, the prospectus reads as both a legal necessity and a statement of genuine conviction, leaving shareholders to decide whether that conviction is reassuring or disqualifying.
Source: techcrunch.com
Peak XV raised its Surge seed investment ceiling to $5 million and announced a new 18-startup cohort, with most targeting global markets, from TechCrunch AI.
Peak XV, the Sequoia Capital spinoff, has bumped the top ticket size for its Surge early-stage program from $3 million to $5 million. The firm unveiled the 18 startups in its latest cohort alongside the increase, signaling a deeper commitment to seeding ambitious founders in South and Southeast Asia.
Thirteen of the 18 startups are building for global markets from day one, while more than half are headquartered in India. The mix spans sectors like AI, developer tools, and cross-border commerce — a tilt toward founders who see their addressable market as the world, not just their home region.
The new ceiling applies to the program’s initial check, which can now reach $5 million per company. Peak XV has run Surge since 2019, positioning it as a fast-track for pre-Series A teams; the larger checks suggest the firm is willing to write bigger bets earlier to win deals in a crowded seed market.
Peak XV did not disclose the total capital committed to this cohort. The announcement lands as Indian startups see a rebound in funding activity, with top-tier investors competing more aggressively for early-stage stakes in AI and software companies.
Source: techcrunch.com
OpenAI has dropped a model due to safety concerns over its poor instruction-following, per a top executive's comments to the Wall Street Journal, from TechCrunch AI.
A senior OpenAI executive told the Wall Street Journal the company scrapped a model over safety issues. The core problem: the model showed a weak aptitude for following orders, which made it unreliable for deployment.
The decision was made internally before any public release. No name for the model was given, and OpenAI has not issued a formal statement beyond the executive's remarks to the Journal.
This move fits a pattern of heightened caution at the lab, which has increasingly stressed alignment work as its models grow more capable. The specific failure mode—poor instruction adherence—undermines the basic contract of a useful AI assistant.
Source: techcrunch.com
AI infrastructure provider Modal Labs is reportedly raising a $750 million round at a $15.75 billion valuation, more than tripling its valuation from four months ago, from TechCrunch AI.
The funding round, first reported by a source familiar with the matter, would value Modal Labs at $15.75 billion. That marks a sharp jump from the company’s valuation in a round that closed roughly four months ago, according to the report. The new investment is expected to more than triple that prior figure.
Modal builds infrastructure for running AI workloads, competing in a crowded field of cloud and compute providers catering to model training and inference. The company has been growing quickly as demand for GPU access and managed AI deployment tools surges.
The round is still pending final terms, the source said. If completed, it would place Modal among the most richly valued AI infrastructure startups, reflecting investor appetite for compute layer bets even as public markets remain volatile.
Source: techcrunch.com
AMD is acquiring AI research lab World Labs, co-founded by Fei-Fei Li, in an all-stock deal valued at about $8.2 billion, from The Verge AI.
AMD announced the acquisition today, paying entirely in stock for World Labs, which launched in 2024 and reached a $1 billion valuation within months. The deal is expected to close by the end of the year. World Labs’ first commercial product, Marble, a world generation model that creates interactive 3D environments from text prompts, shipped in 2025.
Li will join AMD as EVP and chief scientist, reporting directly to CEO Lisa Su. The World Labs team stays intact, focused on advancing AI model research under AMD’s umbrella. The purchase price marks an eightfold jump over the startup’s last known valuation, signaling AMD’s intent to compete aggressively in the AI infrastructure race.
Source: theverge.com
OpenAI is expected to announce a consumer AI agent called Aeon at its DevDay on Tuesday, aiming to catch up in a category where rivals like Meta's Muse and OpenClaw have moved ahead, from The Verge AI.
OpenAI, which defined the modern chatbot era, has reportedly lagged in the race for continuously running consumer AI agents. The company's response, according to widespread rumors, is Aeon — a direct competitor to Meta's Muse, SpaceX's Grok Bot, and the open-source OpenClaw. The announcement is slated for the company's DevDay event this week.
Aeon's success depends on execution. To overtake the field, it would need to combine the best elements of existing platforms: OpenClaw's utility, the affordability of cheaper rivals, and presumably OpenAI's own conversational strengths. The Verge frames this as a pivotal moment — not just a product launch, but an attempt to reclaim leadership in a category OpenAI helped inspire but didn't dominate.
The stakes are straightforward: OpenAI popularized the chatbot, but agents that run continuously and serve consumers are a different game. If Aeon ships with the right mix of capability and price, it could reset the conversation. If not, OpenAI risks watching competitors define the next phase of AI interaction without its involvement.
Source: theverge.com
A nonprofit's breach from a phishing email illustrates how AI tools are making cyberattacks harder for hospitals and banks to defend against, from The Verge AI.
Vivian's Door, an Alabama nonprofit serving minority-owned businesses, had its systems compromised after a phishing email tricked an employee into granting access. Attackers used the foothold to send fraudulent fundraising emails to contacts worldwide, forcing the organization offline for three days while its IT team patched the vulnerability. The incident, which began with phone calls about unsolicited pleas for money, exposed how a single credential lapse can cascade into a reputational and operational crisis.
The Verge's reporting frames this as a warning for critical sectors. Hospitals and banks are increasingly targeted by AI-assisted hacking, which automates reconnaissance, crafts convincing phishing lures, and scales attacks faster than human defenders can respond. Legacy security infrastructure in these institutions—often built around perimeter defenses rather than identity verification—struggles to keep pace with AI's ability to iterate and evade.
For smaller organizations like Vivian's Door, the gap is even starker. They lack the dedicated security teams of large financial firms, making them attractive entry points for attackers seeking to pivot into partner networks. The breach wasn't exotic—just a well-timed phishing email—but the speed and personalization AI enables made it harder to spot until damage was done.
The stakes are concrete: a compromised nonprofit's contact list became a spam relay, but the same technique against a hospital could reroute billing or lock down patient records. The article underscores that AI doesn't create new vulnerabilities—it simply makes old ones cheaper to exploit at scale, and most local institutions haven't updated their defenses accordingly.
Source: theverge.com
Florida's attorney general is asking a court to stop OpenAI from making ChatGPT seem human, following the state's earlier safety lawsuit, from The Verge AI.
James Uthmeier filed a motion seeking a preliminary injunction against OpenAI, arguing ChatGPT's first-person pronouns and simulated emotion deceive users into treating the bot as a trusted friend. The filing claims this anthropomorphism is a deliberate design choice that boosts engagement and feeds OpenAI's training loop, while making the product less reliable as users lean on it more.
The request builds on Florida's existing lawsuit against OpenAI, filed months ago over broader safety concerns. Uthmeier's latest move narrows the focus to the product's conversational interface itself, framing the alleged harm as psychological rather than purely technical—users aren't just getting wrong answers, they're being manipulated into misplaced trust.
No ruling has been issued yet, and OpenAI has not publicly responded to this specific motion. The case will test whether courts treat simulated personality as a deceptive practice under Florida law, a question with ripple effects for any company shipping conversational AI to consumers.
Source: theverge.com
Politics
Jack Smith, Former Special Counsel, Will Appear Before Congress from NYT vs Jack Smith to testify about Trump probes before Senate panel today from CBS News
(NYT, lean-left) Jack Smith, Former Special Counsel, Will Appear Before Congress: Mr. Smith’s testimony will be the latest public clash over the use of prosecutorial power in American politics.
(CBS News, lean-left) Jack Smith to testify about Trump probes before Senate panel today: Former special counsel Jack Smith will testify Tuesday before the Senate Judiciary Committee as part of its investigation into his probes of President Trump.
Source: NYT (lean-left, nytimes.com), CBS News (lean-left, cbsnews.com)
3 news outlets seek extension of ban on their being blocked from White House from CBS News vs News organizations ask court to bar White House enforcement of access ban from NPR
(CBS News, lean-left) 3 news outlets seek extension of ban on their being blocked from White House: CNN, MS NOW and Politico, which President Trump abruptly banished from White House grounds this month, asked a federal court to block the ban until the case against it is resolved.
(NPR, center) News organizations ask court to bar White House enforcement of access ban: Three news outlets that President Trump banished from White House grounds asked a federal court to block the ban until the case is resolved.
Source: CBS News (lean-left, cbsnews.com), NPR (center, npr.org)
Senate approves sweeping bill to regulate college sports from CBS News vs The Senate passes the Protect College Sports Act, but the bill's future is unclear from NPR vs Senate passes Protect College Sports Act to regulate transfers, paying athletes from New York Post
(CBS News, lean-left) Senate approves sweeping bill to regulate college sports: The Senate on Thursday approved a sweeping college sports bill to establish a federal framework for college athletics aimed at protecting student athletes — despite some pushback from some Democrats who argue that it could do the opposite.
(NPR, center) The Senate passes the Protect College Sports Act, but the bill's future is unclear: The bill's supporters, which include the NCAA, say it will tame what they call a "Wild West" era of college sports as litigation chips away at long-held rules around eligibility, transfers and pay.
(New York Post, lean-right) Senate passes Protect College Sports Act to regulate transfers, paying athletes: The Senate passed a sweeping college sports bill Monday to set national standards and protections for athletes after years of failed attempts to wrangle through comprehensive reform.
Source: CBS News (lean-left, cbsnews.com), NPR (center, npr.org), New York Post (lean-right, nypost.com)
Trump administration says it's scaling back fuel efficiency standards for new cars from NPR vs Pete Buttigieg discusses Trump administration's rollback of fuel efficiency standards from NPR
The Trump administration is weakening Biden-era fuel efficiency standards for new cars and trucks.
Source: NPR (center, npr.org), NPR (center, npr.org)
Five arrested as counter-terror police investigate major incident near UK airbase used by US forces from BBC News vs Rubio Says Incident at UK Air Base RAF Fairford ‘Clearly Involved’ a Foreign Actor from NYT
(BBC News, center) Five arrested as counter-terror police investigate major incident near UK airbase used by US forces: The men have been detained on suspicion of preparation of a terrorist act after three suspicious vehicles were reported to be travelling towards the base.
(NYT, lean-left) Rubio Says Incident at UK Air Base RAF Fairford ‘Clearly Involved’ a Foreign Actor: The U.S. secretary of state, Marco Rubio, told Fox News that what happened at a base used by U.S. bombers flying to Iran was “very serious,” without providing details.
Source: BBC News (center, bbc.co.uk), NYT (lean-left, nytimes.com)
Far-left Senate hopeful Abdul El-Sayed asks if football is ‘ethical’ —despite being captain of HS team — in resurfaced podcast from New York Post vs Democrat Abdul El-Sayed courts RFK Jr.'s MAHA voters with new health platform from CBS News
(New York Post, lean-right) Far-left Senate hopeful Abdul El-Sayed asks if football is ‘ethical’ —despite being captain of HS team — in resurfaced podcast: "What is it about this brutal sport that keeps us coming back?" El-Sayed is heard saying. "Is there a version of football that's actually ethical and safe? What would need to change?"
(CBS News, lean-left) Democrat Abdul El-Sayed courts RFK Jr.'s MAHA voters with new health platform: Michigan Democratic Senate candidate Abdul El-Sayed is making a play for Robert F. Kennedy Jr.'s "Make America Healthy Again" supporters with the release Tuesday of a new health policy platform.
Source: New York Post (lean-right, nypost.com), CBS News (lean-left, cbsnews.com)
Mamdani allies hosting fundraiser for Dem Senate hopefuls James Talarico, Sherrod Brown from New York Post vs Top Democrats Start to Fund James Talarico in Texas Senate Race from NYT
(New York Post, lean-right) Mamdani allies hosting fundraiser for Dem Senate hopefuls James Talarico, Sherrod Brown: Allies of Mayor Zohran Mamdani are hosting a fundraiser Monday for Senate Democratic hopefuls James Talarico of Texas and Sherrod Brown of Ohio — including one who referred to America as a "s—thole."
(NYT, lean-left) Top Democrats Start to Fund James Talarico in Texas Senate Race: The main super PAC for Senate Democrats will begin to run digital ads in the state’s Senate race, in a sign of the party’s growing confidence.
Source: New York Post (lean-right, nypost.com), NYT (lean-left, nytimes.com)
A Tough Ad Reminds Nebraska Voters About a Brutal Crime and a Pardon from NYT
A group supporting Dan Osborn, the independent Senate candidate in Nebraska, is airing an ad attacking Senator Pete Ricketts, a Republican, for pardoning a man convicted of sexual assault.
Source: NYT (lean-left, nytimes.com)
Democratic Turnout Surged in the Primaries. Will It Matter in November? from NYT
A voter enthusiasm gap provides another signal that Democrats are favored this fall, though Republicans still have key advantages.
Source: NYT (lean-left, nytimes.com)
U.S. ban on almost $1 billion in Canadian imports takes effect from CBS News
The United States has banned nearly $1 billion worth of Canadian imports, including alcoholic beverages, dairy products and motorcycles.
Source: CBS News (lean-left, cbsnews.com)
Trump denies offering Iran sanctions relief for nuclear concessions – US politics live from The Guardian
President uses social media post to deny reports that he was willing to ease sanctions in return for ‘concrete’ steps over its nuclear program
Sign up to the US Breaking News email
Donald Trump will host a meeting today with House speaker Mike Johnson and tech executives like Meta CEO Mark Zuckerberg and Anthropic’s Dario Amodei to discuss AI, as calls ramp up for more regulation of the fast-developing technology.
Nvidia CEO Jensen Huang and OpenAI president Greg Brockman are also expected to attend, separate sources familiar with the planning told Reuters.
Continue reading...
Source: The Guardian (lean-left, theguardian.com)
Three deaths, three states, 24 hours: one day in ICE’s quiet but ‘gruesome’ surge from The Guardian
These men in Maine, Georgia and Florida shared no ties – until they became casualties in the latest phase of Trump’s anti-immigration campaign
Gone in 90 seconds: an ‘invisibilized’ ICE is arresting more people than ever
It was 7am on 13 July this year. A warm, sunny morning, as residents of Biddeford, Maine, left their homes, got into their cars and began their commute. Joan Sebastián Durán Guerrero came out of the apartment he lived in with his partner and daughter and drove off in his white car. Minutes later, neighbors heard three pops. Durán Guerrero was dead.
More than 1,200 miles (1,900km) away in Ocilla, Georgia, Jesús Manuel Arenas-Silva was shuffled on to a bus. Private sector guards and Immigration and Customs Enforcement (ICE) officials were beginning the process of transferring him and several other immigrants from one detention center to another nearly two hours away. Ten minutes into the ride, he became unconscious. Doctors at a nearby hospital pronounced him dead.
Continue reading...
Source: The Guardian (lean-left, theguardian.com)
Samuel Alito steps aside in a major climate case amid scrutiny over oil stock holdings from The Guardian
Decision from conservative justice comes before court hears case on whether fossil-fuel firms deceived US public
Supreme court justice Samuel Alito recused himself on Monday from a major climate-change case after facing calls to step aside due to stock holdings in oil companies.
A letter posted in the case did not specify a reason, but the conservative justice has previously recused himself from other cases involving companies where he holds stock.
Continue reading...
Source: The Guardian (lean-left, theguardian.com)
Trump announces $15bn Iowa steel plant as he bids to boost flagging midterm hopes from The Guardian
President says plant will be built by Indian-owned company and calls it largest such facility in American history
Donald Trump on Monday announced what he called the largest steel plant in American history to be built in Iowa, giving Republicans a jobs message to take to voters in a state that could help decide control of the Senate in November.
“Using advanced steelmaking technology, the plant will produce some of the highest quality, most affordable steel anywhere in the world,” Trump said in the Oval Office, flanked by Iowa’s top Republicans.
Continue reading...
Source: The Guardian (lean-left, theguardian.com)
Trump appointee overseeing election investigations steps down from FBI role from The Guardian
Andrew Bailey will resign as FBI co-deputy director about a year after assuming the position
Andrew Bailey, a deputy director of the FBI who has been overseeing some of Trump’s election-related investigations, will step down from his position approximately one year after assuming his role.
“I am announcing that I am stepping down to return to my family in Missouri,” Bailey wrote in an X post on Monday amid an ongoing shake-up at the bureau.
Continue reading...
Source: The Guardian (lean-left, theguardian.com)
Could a far left candidate win a midterm race in a red state like Florida? from NPR
Angie Nixon, the Democratic U.S. Senate Candidate in Florida, is running from the far left — raising questions about whether such a campaign can work in a red state like Florida.
Source: NPR (center, npr.org)
US prosecutors reopen case of alleged gang rape at Cornell University from BBC News
It comes after the woman filed a civil lawsuit alleging she was drugged and raped at a fraternity house in 2024.
Source: BBC News (center, bbc.co.uk)
Tennessee governor declines to halt execution of state's lone woman on death row from BBC News
If the execution proceeds, Pike would be the first woman put to death in Tennessee in more than 200 years.
Source: BBC News (center, bbc.co.uk)
Watch: From launch to fiery splashdown, SpaceX's Starship completes orbit test from BBC News
A thunderous lift off, rapid satellite deployments and a fiery ending. That's how SpaceX's 14th launch of its Starship rocket panned out today.
Source: BBC News (center, bbc.co.uk)
New York Times executive fatally shot by elderly in-laws, police say from BBC News
Jonathan McKinsey's in-laws, who are both 77 years old, face multiple charges including first-degree murder.
Source: BBC News (center, bbc.co.uk)
Trump, China and the question of selling arms to Beijing from The Hill
Questions about whether the U.S. could change its posture toward China are on the rise after David Perdue, the U.S. ambassador to China, made waves by saying President Trump had asked Chinese President Xi Jinping if he would “like to buy some.” The president on Monday denied making the offer to Xi, while close observers...
Source: The Hill (center, thehill.com)
Tariff challengers take Trump back to court: 'Rinse and repeat' from The Hill
President Trump hopes the third time will be the charm as his latest round of tariffs heads to court this week. The Supreme Court already ruled Trump can’t invoke an emergency statute to impose his global trade war. Now, the president is pointing to a different law, and this time, everyone agrees it allows the president to impose...
Source: The Hill (center, thehill.com)
Council of Europe head revives Biden-backed framework for AI regulation from The Hill
The Biden administration and a number of countries in Europe have endorsed a framework for AI regulation that respects human rights, democracy and the rule of law, but it is generally unheard of outside of the tiny European organization that drafted it. Alain Berset is the secretary-general of the Council of Europe, an important but...
Source: The Hill (center, thehill.com)
Cuba tensions on the rise as Trump pushes for 'fundamental change' from The Hill
Tensions between the U.S. and Cuba are rising once again after President Trump told world leaders gathered at the United Nations General Assembly that he was seeking “fundamental change” from the island and asserting that “freedom will be coming to Cuba.” The Trump administration has never walked back threats of a possible military intervention —...
Source: The Hill (center, thehill.com)
GOP senators push back on Trump’s taxpayer-funded ad from The Hill
President Trump's new government-funded ads pledging to “cast out the communists, Marxists and fascists” are getting strong pushback from Senate Republicans who are warning the president not to use taxpayer money to promote himself in such a brazenly political way. While Republicans generally approve of Trump’s political message, they say using taxpayer dollars to fund...
Source: The Hill (center, thehill.com)
Could Trump revoke Ilhan Omar's citizenship for alleged brother marriage? Ex federal prosecutor breaks it down from Fox News
DHS is conducting a criminal investigation into Rep. Ilhan Omar, but a former federal prosecutor warns of "very high bars" to prove marriage fraud.
Source: Fox News (lean-right, foxnews.com)
Socialist jeweler who called US a 'sh--hole' teams up with top Dem Senate hopefuls from Fox News
Alexis Bittar, who called America a "sh--thole" and the American flag a symbol of aggression, hosts a fundraiser in NYC supporting Democratic Party candidates for U.S. Senate.
Source: Fox News (lean-right, foxnews.com)
New audit in Walz's backyard exposes massive fraud risk tied to post-George Floyd activism from Fox News
A Minneapolis audit reveals weak oversight of $36 million in violence prevention contracts tied to programs launched after George Floyd's death.
Source: Fox News (lean-right, foxnews.com)
Hero K-9 sniffs out hidden $6M cocaine shipment along US-Canada border from Fox News
Customs and Border Protection officers and a canine unit uncovered 174 vacuum-sealed cocaine packages hidden in a tractor-trailer's modified floor.
Source: Fox News (lean-right, foxnews.com)
State Department considers sharing passport records for voter citizenship checks from Fox News
The State Department is assessing how to share passport records with state and local officials to verify citizenship on voter registration rolls.
Source: Fox News (lean-right, foxnews.com)
DHS watchdog finds vetting errors in 40% of files for migrants who entered US during Biden’s botched Afghanistan withdrawal from New York Post
In two cases, missed aliases and incomplete documentation involved individuals who may have potentially been inadmissible to the US on “terrorism-related grounds.”
Source: New York Post (lean-right, nypost.com)
Trump administration’s Department of Education scraps Biden-era Title IX rules from New York Post
"We will continue to relentlessly champion equal opportunity for all Americans and hold accountable any school or college that violates the rights, privacy, or athletic opportunities of our women and girls," said US Secretary of Education Linda McMahon.
Source: New York Post (lean-right, nypost.com)
A captivating new biography of the flawed fixer Roy Cohn from Washington Examiner
When Roy Cohn died of AIDS in 1986, Donald Trump was among the few to attend his burial at Union Field Cemetery in Queens. He also attended Cohn’s public memorial later that year in Manhattan, along with Rupert Murdoch, George Steinbrenner, Barbara Walters, and William F. Buckley. Apart from his legions of well-connected friends and […]
Source: Washington Examiner (lean-right, washingtonexaminer.com)
And just like that, Germany’s establishment went kaput from Washington Examiner
The British diplomat and novelist, Harold Nicholson, tossed off an essay in 1929 containing a short riff on the three great capitals of Europe. “London is an old lady in black lace and diamonds who guards her secrets with dignity,” while “Paris is the woman in the prime of life to whom one would only […]
Source: Washington Examiner (lean-right, washingtonexaminer.com)
Call them ‘Nazis’ all you want: The new German Right is winning the argument from Washington Examiner
In the current electoral cycle, Germans of a particular political persuasion can vote for the slightly left-of-center Social Democrats (SPD), the even more left-of-center Greens, the straightforward left that calls itself The Left party (Die Linke), the far-left Sahra Wagenknecht Alliance (BSW), or the decayed remains of the former East German Communist Party. But for […]
Source: Washington Examiner (lean-right, washingtonexaminer.com)
Have Americans forgotten how to argue? from Washington Examiner
Americans disagree about almost everything. We argue about very little. The distinction matters. Disagreement is easy. I state my position. You state yours. We discover that they are incompatible. If we are sufficiently invested in the dispute, we repeat ourselves, usually with greater confidence. Eventually, we may stop talking about the issue and start talking […]
Source: Washington Examiner (lean-right, washingtonexaminer.com)
Republicans’ chances at holding on to Senate ‘looking good’: Sarah Bedford from Washington Examiner
Washington Examiner investigations editor Sarah Bedford said Republicans may still have a good chance at holding the Senate. “It’s going to be tough, but I think it’s looking good,” Bedford said on The Hugh Hewitt Show on Wednesday. “They’ve got some pickup opportunities in Michigan, for example.” The Michigan Senate race between Democratic nominee Abdul […]
Source: Washington Examiner (lean-right, washingtonexaminer.com)
Which Books Are Best On Women’s Rights In Asia In WWII? A Reading Guide To War, Colonialism And Historical Memory from Bitrebels.com
Anyone looking for books about women’s rights in Asia during the Second World War will quickly encounter a subject that is larger than the wartime battlefield. Questions about sexual violence, military prostitution, colonial rule, women’s labor, and individua…
Source: Bitrebels.com (unrated, bitrebels.com)
Our closing argument: The midterms are a referendum on who runs your life from Fox News
House Republicans frame the 2026 midterms as a stark fight between individual liberty and democratic socialism's push for government control.
Source: Fox News (lean-right, foxnews.com)
Lindsay Clancy holdout juror’s lawyer challenges claim he broke the law and more top headlines from Fox News
Get all the stories you need-to-know from the most powerful name in news delivered first thing every morning to your inbox.
Source: Fox News (lean-right, foxnews.com)
UK voters believe Burnham can win the next election from POLITICO.eu
Voters think the new prime minister has transformed Labour’s chances, but doubt he will help with the cost of living, according to Public First research.
Source: POLITICO.eu (unrated, politico.eu)
Erik Visits an American Grave, Part 2,240 from Lawyersgunsmoneyblog.com
This is the grave of James Inhofe. This is an unusual grave post because it’s one of only a few times that I’ve visited the grave of someone who I always wrote an obituary about. It does make it pretty easy for me though because then I can just republish the …
Source: Lawyersgunsmoneyblog.com (unrated, lawyersgunsmoneyblog.com)
How You’re Secretly Hurting, Based On Your Birth Month from Thoughtcatalog.com
April You need help, but won’t ask for it. Responsibility is weighing heavy on your shoulders, but you see this task as your obligation, so you keep your mouth shut and plod on. The problem is the …
Source: Thoughtcatalog.com (unrated, thoughtcatalog.com)
Links 9/28/2026 from Nakedcapitalism.com
Our confident daily links: Trump hawks US weaponry to China, RAF base bomb plot, fake rich people, and Melania defends communism.
Source: Nakedcapitalism.com (unrated, nakedcapitalism.com)
Voters Back Democrats But Still Don’t Trust Them from Politicalwire.com
Playbook: “A new polling memo from Navigator Research being circulated to senior Hill leadership offices this morning is arguing Democrats’ affordability agenda is popular with over 70% of voters they polled, including 55% of MAGA Republicans. And the memo la…
Source: Politicalwire.com (unrated, politicalwire.com)
High Potential Season 3 in The Daily LITG, 28th September 2026 from Bleeding Cool News
High Potential Season 3 was the most-read article on Bleeding Cool yesterday. Lying In The Gutters is the daily runaround for the most-read stories the day before, as well as over the past seven years. Founded seventeen years ago and steeped in a history of c…
Source: Bleeding Cool News (unrated, bleedingcool.com)
Wife school can’t fix MAGA husbands from Salon
A new wave of Christian influencers claim women can save their marriages with more self-loathing
Source: Salon (unrated, salon.com)
Science News
A new device captures CO2 by pumping it across a battery, using less energy than current systems, from Ars Technica Science.
Researchers built an electrochemical cell that pulls carbon dioxide from a gas stream by cycling it through a battery-like setup. The device relies on a chemical reaction that binds CO2 during one phase and releases it during another, letting the gas be collected separately without high heat or pressure.
The system’s key advantage is energy efficiency: it operates at near-ambient conditions and uses voltage swings to drive capture and release, avoiding the thermal costs typical of amine-based scrubbers. In testing, the device maintained stable performance over multiple cycles, though the lab-scale prototype still needs scaling to handle industrial flue-gas volumes.
The work points toward a simpler pathway for post-combustion capture, but practical deployment depends on improving electrode longevity and reducing material costs. The authors emphasize that the battery-like architecture could integrate with renewable electricity, making the process easier to power cleanly, though no pilot-scale results were reported.
Source: arstechnica.com
Scientists corrected a typo that had misdated an 1841 geomagnetic storm account, resolving a 1840s space weather mystery from Ars Technica Science.
The original mystery stemmed from a famous 19th-century Nature article describing a train delayed by a geomagnetic storm in 1841. But the event didn’t match known solar activity records for that year, leaving historians puzzled for decades. The new analysis shows the date was a typo — the storm actually occurred in 1842, aligning with a documented sunspot maximum.
The correction came from re-examining the original handwritten records and cross-referencing them with telegraph and railway logs from the period. The 1842 storm had already been linked to auroras visible at unusually low latitudes, and the train delay fits that timeline perfectly. The typo likely crept in during the original transcription or typesetting process.
By fixing the year, the study also strengthens the case that geomagnetic storms were disrupting early electrical and telegraph systems more frequently than previously assumed. The 1842 event now serves as a clearer benchmark for how mid-19th-century infrastructure responded to extreme space weather. It also highlights how small archival errors can ripple through scientific literature for over a century.
Source: arstechnica.com
US solar growth is expected to continue outpacing a 2% rise in electricity demand in 2026, per Ars Technica Science.
The article reports that renewable capacity additions, led by solar, are on track to comfortably cover the projected 2% increase in US electricity demand for 2026. This means the grid’s new demand will be met without a corresponding rise in fossil-fuel generation, a shift from past years when gas often filled the gap.
Data from the US Energy Information Administration shows solar installations are still climbing, though the pace may be slightly less explosive than the record years. Supply chain bottlenecks and interconnection queue delays have moderated the growth rate, but the baseline remains strong — enough to absorb demand growth and then some.
The practical effect: utilities and grid operators are less likely to need emergency peaker plants or extended coal runs during summer peaks. Instead, the margin between supply and demand is widening, even as electrification (EVs, heat pumps, data centers) pushes consumption upward.
That said, the article cautions that the 2% figure is a national average; regional grids — especially in the Southeast and parts of the Midwest — still face tighter conditions if solar buildout lags local load growth. The national cushion doesn’t erase local pinch points.
Source: arstechnica.com
Scientists found that ice grains escaping Saturn's moon Enceladus may concentrate and separate ocean chemicals as they freeze and shatter, offering a new explanation for Cassini's ice-grain data from Science Daily.
Researchers studying Cassini's observations of Enceladus' plumes now think the icy particles undergo a natural purification process. As droplets from the subsurface ocean freeze and fragment during their high-speed ejection, they can selectively trap or exclude certain compounds, sorting chemicals by their solubility and freezing behavior.
That sorting effect likely skews what Cassini detected in the plume grains, meaning the spacecraft's readings were not a straightforward sample of the ocean's bulk composition. The team's models show the process could amplify the presence of some organic molecules and diminish others, which would explain why certain chemical signatures appeared stronger or weaker than expected.
The finding also points to a practical opportunity: future missions could exploit this separation mechanism. By targeting the smallest or most processed ice grains, spacecraft might pull out trace organic compounds that would otherwise remain undetectable in the ocean's diluted mix. The paper's authors suggest this could sharpen the search for biosignatures without needing to drill through miles of ice.
Source: sciencedaily.com
Scientists may have solved why Earth’s day length shifts by milliseconds over decades, pointing to inner core dynamics, from Science Daily.
The article reports a proposed explanation for a long-observed quirk: Earth’s rotation isn’t perfectly steady, and day length can vary by a few milliseconds across decades. The new work suggests the cause lies in the planet’s inner core, whose subtle movements appear to tug on the mantle above it. That interaction would slightly alter the planet’s angular momentum, stretching or shrinking the day.
The inner core, long treated as a static, solid sphere, is shown to be more dynamic than assumed. Small shifts in its position or orientation could couple with the outer core’s magnetic field and the mantle’s mass distribution, producing measurable rotational wobbles. This isn’t a daily tick—the changes accumulate over years to decades, matching the observed pattern of day-length drift.
Researchers aren’t claiming a fully closed case. The model ties existing seismic and geomagnetic data to rotational changes, but direct observation of the core remains impossible. Instead, they infer behavior from surface signals, a standard approach in deep-Earth science. The result is a working hypothesis that links the planet’s interior to its surface timekeeping.
Source: sciencedaily.com
NASA and SpaceX could launch Crew-13 to the ISS on October 1, with a Dragon trip of under nine hours from Florida, per planning details from Science Daily.
The four-person Crew-13 mission is targeting liftoff on October 1, with the Dragon spacecraft expected to dock at the International Space Station less than nine hours after launch. That fast profile, if it holds, would put the crew aboard the orbiting lab the same day they leave the pad.
The flight would join the crew of Expedition 75 once on orbit. SpaceX and NASA are still working through a final round of launch-readiness checks before committing to the date, so the schedule remains contingent on those reviews going clean.
No crew members have been named in the release, and the specific launch window from Florida hasn’t been detailed. The short transit time is notable only because it cuts a typical multi-day journey down to a single work shift, which has become standard for recent Dragon missions.
Source: sciencedaily.com
Gaming News
Dan Trachtenberg is developing a Sleeping Beauty-inspired movie for Disney, leaving Predator's future unclear from IGN.
IGN reports Trachtenberg, who directed 2022's *Prey*, is now attached to a new Disney project based on Sleeping Beauty. The article doesn't specify whether it's live-action or animated, nor does it name a writer or producer.
The news raises questions about the Predator franchise's next installment. *Prey* was a critical and commercial hit, but no sequel has been officially greenlit. Trachtenberg's Disney commitment doesn't confirm a Predator follow-up is dead—it just means his immediate focus has shifted.
Disney has been mining its animated classics for live-action and reimagined features, and Sleeping Beauty joins a slate that already includes *Snow White* and *Moana* adaptations. Trachtenberg's genre background suggests this could be a darker take, but IGN's report offers no plot details or timeline.
Source: ign.com
The TARDIS set from Doctor Who has been dismantled as the show remains on hiatus, from IGN.
The iconic interior of the TARDIS — the police box that serves as the Doctor's time machine — has been taken down. The removal coincides with the ongoing production pause for the long-running BBC series, which has not been actively filming new episodes.
No official reason was given for the teardown, and it's unclear whether the set is being permanently retired or simply stored pending the show's eventual return. The structure had been a fixture of the series since its 2005 revival, undergoing several redesigns over the years.
The news surfaces as fans wait for updates on the next season, with no confirmed production timeline announced. The set's removal adds to the uncertainty surrounding the show's immediate future, though it does not necessarily signal cancellation.
Source: ign.com
Alan Cumming says he didn't fully understand Avengers: Doomsday's plot while filming, including a scene where he had to lie on top of Florence Pugh's Black Widow, from IGN.
Cumming, who reprises his X-Men role as Nightcrawler in the upcoming Avengers film, told IGN he was kept in the dark about the larger story. "I don't fully understand the plot," he said, adding that the secrecy extended to his own scenes. He described one sequence where he was required to lie on top of Pugh's character, noting the physical direction was explicit but the narrative context was not.
The actor said he filmed his part without a full script, a common practice for Marvel's blockbuster productions, which often shoot in fragments to prevent leaks. Cumming joked that he simply followed instructions, trusting the studio's vision even as he remained puzzled about how his character fit into the multiversal chaos. He did not specify how long his shoot lasted or whether he interacted with other cast members beyond Pugh.
Cumming's comments echo previous remarks from Marvel actors who have admitted to confusion during production, though his admission stands out for its candor. Doomsday is slated for a 2026 release, and the film is expected to feature a sprawling cast of returning and new heroes. The actor's lack of clarity suggests even key players are working with limited information, a testament to Marvel's ongoing effort to guard plot details.
Source: ign.com
Control’s resonant director explains the deliberate omission of yellow paint as a navigation aid, favoring environmental storytelling from Kotaku.
In a recent interview, the game’s resonant director confirmed that the absence of yellow paint—the industry-standard visual cue for climbable or interactive surfaces—was a conscious design choice. The team wanted players to engage with the world’s architecture and lore rather than follow color-coded breadcrumbs. He stressed that the Brutalist setting of the Oldest House was meant to be read like a puzzle, not a checklist.
The decision was tested internally, with some playtesters initially frustrated. But the director argued that the confusion is part of the intended experience: “If you want to get the most out of this world, you have to participate.” The game’s environmental artists compensated by embedding visual logic into lighting, signage, and material contrast, rewarding players who studied their surroundings.
This stance sets Control apart from many modern triple-A titles that lean on overt wayfinding aids. The director acknowledged the risk, noting that not every player will appreciate the extra cognitive load. But he stood by the choice, framing it as a bet on the player’s curiosity over convenience—a gamble that evidently paid off for the game’s cult following.
Source: kotaku.com
Avengers: Endgame's re-release included a bonus Doomsday tease only visible in IMAX screenings, not regular showings, from Kotaku.
Marvel's Avengers: Endgame returned to theaters with four new post-credit scenes, but only IMAX audiences saw all of them. Regular theaters received just three of the four, with the fourth—a brief shot teasing the arrival of Doomsday—exclusive to the premium large-format experience. The discrepancy wasn't advertised upfront, leaving standard-format viewers to discover the missing scene only after the credits rolled.
The scene in question is a short, wordless beat that connects directly to the film's broader narrative threads, but its placement as an IMAX-only extra means most ticket buyers never encountered it. Marvel and Disney did not issue a clarification before the re-release, and the exclusion only surfaced through audience reports comparing notes online. Whether the omission was a technical decision or a deliberate incentive to push IMAX ticket sales remains unconfirmed, but the studio has not disputed the reports.
For those who caught the encore in IMAX, the Doomsday tease adds a layer of foreshadowing that regular viewers now have to track down secondhand. The re-release itself was framed as a thank-you to fans, but the uneven distribution of new content undercuts that message—especially since the missing scene isn't a throwaway gag but a plot-relevant hint. No official statement has explained why the scene was held back from standard formats.
Source: kotaku.com
A Sleeping Beauty movie is reportedly in development from the director behind the recent Predator revival, according to a report from Kotaku.
The report, sourced from industry insiders, claims director Dan Trachtenberg — best known for helming 2022’s *Prey*, which revived the Predator franchise — is attached to a new live-action *Sleeping Beauty* project for Disney. No casting, plot details, or release window have been confirmed, and Disney has not publicly commented on the report.
The project would mark Trachtenberg’s first major studio fairy tale, a notable pivot from his sci-fi and horror work. It also fits Disney’s ongoing push to reimagine its animated classics, though *Sleeping Beauty* has seen less recent attention than *Cinderella*, *Beauty and the Beast*, or *The Little Mermaid*.
If the deal goes through, Trachtenberg would bring a lean, grounded action sensibility to a property known for its ornate visuals and passive princess — a combination that could either refresh the formula or clash with Disney’s established tone. As with any unconfirmed report, the project’s status may shift before an official announcement.
Source: kotaku.com
