OpenAI announced on Thursday a new family of AI models, GPT-5.6, comprising three versions — Sol, Terra and Luna — and made them available through ChatGPT, Codex and the OpenAI API.

The lineup expands the company’s offerings in enterprise work, coding and scientific research. Sam Altman said the models are substantially more efficient and less costly than earlier versions, noting that Sol achieves 54 % greater token efficiency for AI coding tasks.

OpenAI describes GPT-5.6 as its strongest cybersecurity model, capable of threat modeling, code review, patching and blue‑team exercises. The Trump administration previously sought to limit the model’s distribution, citing concerns over potential misuse. The new ChatGPT Work tool was also released, offering enterprise teams a desktop, web and mobile companion for drafting documents, spreadsheets and presentations.

The release follows similar announcements by SpaceXAI and Meta earlier this week. OpenAI positioned the models as competitive with Anthropic’s offerings, citing the Artificial Analysis Coding Agent Index, which it says shows Sol scoring 80, 2.8 points higher than Anthropic’s Fable 5, while using less than half the output tokens and costing about one‑third less. Terra scores slightly above Fable 5, and Luna outperforms Opus 4.8, according to OpenAI.

Pricing details show Sol costs $5 per million input tokens and $30 per million output tokens, Terra $2.50 input / $15 output, and Luna $1 input / $6 output.

Wally Funk, an American aviator and spaceflight participant, died on July 8 at age 87 in Grapevine, Texas, after a career that spanned six decades of aviation and a historic suborbital spaceflight in July 2021.

She earned her pilot’s license as a teenager and joined a private astronaut training program in 1961 that tested female pilots for potential NASA selection.

Funk was among the 13 women known as the Mercury 13, a group that underwent physiological tests similar to those of the Mercury Seven astronauts but was never selected for spaceflight.

She later became the Federal Aviation Administration’s first female safety inspector in 1971 and served as one of the National Transportation Safety Board’s early air safety investigators.

In July 2021, at age 82, she flew aboard Blue Origin’s New Shepard rocket as an honored guest, reaching an altitude of more than 60 miles and experiencing a few minutes of weightlessness before landing in West Texas.

Her flight made her the oldest person to reach space at that time, a record later broken by William Shatner and Ed Dwight.

The capsule carried Funk, Blue Origin founder Jeff Bezos, his brother, and a Dutch teenager more than 60 miles above the ground after launching from a desert site in West Texas.

She logged 19,600 hours of flight time and taught more than 3,000 people to fly, according to Blue Origin’s announcement.

Funk said after the flight that she wanted to go again, and in an Instagram video posted by Bezos she remarked that nothing had prevented her from pursuing the goal.

Her interest in space dated to a 1960 Life magazine article about pilot Jerrie Cobb, which led her to apply for the women’s astronaut testing program.

The tests included immersion in ice water, ingestion of a pint of radioactive water, isolation in a temperature‑controlled tank, and centrifuge and high‑altitude chamber exercises.

Beyond aviation safety, she competed in air races, worked as a civilian flight instructor, and continued to seek higher altitudes through personal training.

The city of Grapevine announced her death on July 8 but did not disclose a cause.

Funk’s journey from a 1960s test program to a 2021 suborbital flight illustrates a long‑term effort to expand access to space, and her record was later surpassed by other older travelers.

On July 9, 2026, the European Parliament voted to reject a proposal that would have blocked the interim Chat Control rule, which would permit online services to scan user chats for child sexual abuse material, missing the 360‑seat threshold required to stop its reintroduction. The interim rule, introduced in August 2021 as a derogation from the ePrivacy Directive, expired on April 3, 2026, and was set to be revived unless the Parliament achieved a majority exceeding the Council’s position. 314 MEPs voted to scrap the rule while 276 voted to keep it, falling short of the 360 votes needed; a separate amendment to limit scanning to judicially identified accounts also failed. Parliament did secure a majority to exempt end‑to‑end encrypted platforms from the scanning requirement, though the practical impact may be limited. The Council must decide within three months whether to accept the amendments, with a possible conciliation committee if agreement is not reached; approval would keep the rule in force until 2028 or until a permanent child protection regulation is adopted. Former MEP Patrick Breyer described the outcome as undermining democratic will and warned that the rule could enable suspicionless mass surveillance. The debate continues as trilogue negotiations on the permanent Chat Control 2.0 regulation remain unresolved, leaving the balance between privacy rights and law enforcement needs uncertain.

The United States Army has optimized sustainment for permissive environments, but its current logistics model may be insufficient for large‑scale, contested conflict, according to senior Army officials. History shows that operations without adequate sustainment can fail, as seen in the 1941 German invasion of the Soviet Union and in recent fighting in Ukraine, where pervasive sensing and drone attacks have targeted supply routes, according to analysts and open‑source reports. The Army consumes tens of thousands of gallons of fuel daily for an armored brigade, requiring heavy tactical vehicles that are detectable by thermal and electromagnetic signatures, officials said. Current distribution platforms lack the protection needed to survive deep‑strike attacks, and maintenance shortfalls reduce available capacity, according to a recent Army study. Ammunition expenditure rates observed in Ukraine are comparable to those of World War II, creating challenges for transporting heavy 155‑mm shells and rocket pods across contested terrain, the study noted. Centralized supply depots represent high‑value targets for long‑range precision fires, limiting the speed and volume of resupply, analysts said. To survive contested environments, the Army intends to replace its hub‑and‑spoke model with a decentralized network of smaller, mobile sustainment nodes, officials said. Distributed caches of fuel, water, and ammunition, along with autonomous ground vehicles and heavy‑lift drones, are being pursued to reduce reliance on fixed depots, the Army stated. Sustainment units will be equipped with embedded counter‑UAS systems and short‑range air defense to protect themselves, and will be trained to operate in GPS‑denied, signature‑managed conditions, the Army added. The Army’s modernization priorities have historically favored maneuver and firepower over sustainment, a trend reflected in budget allocations and programmatic focus, officials said. Elevating logistics to a primary warfighting function will require reallocating resources, revising the tooth‑to‑tail ratio concept, and incorporating realistic logistical stress tests into training, according to senior leaders. Without improvements in physical resilience, dispersion, and endurance, the Army risks fielding a force with tactical superiority but operational vulnerability, officials warned. Future large‑scale conflict may be decided by the ability of sustainment to endure persistent attack rather than by the speed or range of individual weapons systems.

SpaceX’s Starship rocket can lift more than 100 metric tons to low‑Earth orbit, a capacity that exceeds the Falcon 9’s 27‑satellite limit and enables up to 60 Starlink V3 satellites per launch. The rocket’s Pez‑dispenser deployment system releases satellites one at a time through a side door, eliminating the need for a large fairing. Satellite manufacturers such as Muon Space, Apex, and Terran Orbital are redesigning flat‑panel “FlatSat” buses to match Starship’s capacity, while Amazon’s LEO constellation continues to use conventional designs. An Aerospace Corporation report estimates that reusable Starship flights could reduce launch cost to $67‑$233 per kilogram after multiple uses, compared with several thousand dollars per kilogram for the Falcon 9. NASA and the U.S. military are evaluating Starship for lunar missions and cargo delivery to remote theaters, and scientists propose using its volume for large space telescopes. Competition from Blue Origin’s New Glenn, which also aims at super‑heavy lift, is expected to pressure pricing and accelerate industry adoption. While Starship remains in an experimental phase, its potential to lower costs and expand launch options is driving a shift in satellite architecture and market expectations.

AMD announced that its next‑generation Zen 6 EPYC “Venice” processors will be introduced at the Advancing AI event on July 22‑23. The company’s CTO and EVP, Mark Papermaster, confirmed the schedule in an interview. The Venice architecture is the sixth generation of AMD’s EPYC server line and will be the first computing product to reach production on a 2nm process node. Manufacturing has begun in Taiwan, with additional production planned at TSMC’s Arizona facility. The flagship Venice chip will contain up to 256 cores, a 33 % increase over the current 192‑core Turin lineup, and is projected to deliver more than 70 % higher performance and efficiency than the Zen 5‑based EPYC processors. The design uses the SP7 socket, supports 16‑channel memory with bandwidth up to 1.6 TB/s, and incorporates PCIe Gen 6.0 for improved CPU‑to‑GPU communication. AMD noted that the new EPYC chips are optimized for standalone x86 traditional workloads, addressing the long‑standing x86 install base used by enterprises. While the announcement focuses on the server market, AMD indicated that mainstream Zen 6 client processors may not be released until later this year, with a possible introduction at CES 2027. Papermaster said that recent enterprise discussions highlight the continued relevance of x86 architecture, and that AMD’s sixth‑generation CPUs build on the performance gains achieved since the 2017 Zen launch.

Google announced at its I/O conference that the SynthID watermark verification feature will be integrated into Chrome and Search, allowing users to check whether images carry SynthID markers without uploading them to the Gemini app. The update will also let Chrome verify C2PA content credentials embedded in images, audio, and video files.

SynthID is an invisible watermark applied by Google AI models to indicate origin, while C2PA is a standard for embedding provenance metadata at creation. Both aim to make AI‑generated content identifiable. Their effectiveness depends on broad adoption by AI developers and platforms.

OpenAI said it will embed SynthID into images generated by ChatGPT, Codex, and its API, and already includes C2PA metadata, though such metadata can be stripped when content is shared on social media. Google noted that C2PA can be removed accidentally or intentionally, and many platforms currently strip metadata from uploaded files. Meta plans to use C2PA to label images captured by its cameras on Instagram, which could help users distinguish AI‑generated photos from authentic ones, though specifics about supported devices remain unclear.

Verification tools are limited to content that already contains a watermark or provenance tag, and many open‑source deepfake models do not incorporate these systems. Fact‑checkers have cited SynthID in debunking deepfakes, but widespread use remains limited.

The expanded integration places verification capabilities in front of a large user base, but success will depend on continued adoption by AI providers and platforms, and on whether labeling alone can reliably address deepfake misinformation.

Patreon announced a partnership with Cloudflare to block AI crawlers from accessing creator content. Cloudflare began default blocking of AI crawlers last year and introduced options for publishers to fine‑tune scraper permissions this year. Patreon CEO Jack Conte said the collaboration aims to stop AI models from training on creators’ work without permission. The partnership leverages Cloudflare’s security infrastructure to identify and block automated scrapers. Patreon joins other platforms seeking to control data used for AI training as the technology expands. The action reflects growing concerns over AI data use and may influence how other platforms handle crawler traffic.

Mercor is in talks to raise a round that would value the AI training startup at $20 billion, according to Bloomberg. The discussions are at an early stage, though Mercor has reportedly provided a term sheet at that valuation.

The company previously raised $350 million in a Series C in October at a $10 billion valuation. Founder‑CEO Brendan Foody posted on X that annualized revenue had reached $2 billion, a 100 % increase over four months.

Mercor announced it is acquiring Deeptune, a firm that trains AI agents, with the Deeptune team joining Mercor as part of the deal. Earlier in 2026 the company suffered a data breach and faced lawsuits from contract workers, Business Insider reported.

The potential funding round and acquisition indicate Mercor may be moving past earlier setbacks.

Apple is testing DRAM chips made by CXMT, a Chinese memory company on the US blacklist, for devices sold in China, according to people familiar with the matter. The testing follows Apple’s lobbying effort to have US restrictions on Chinese memory suppliers eased, citing rising memory prices and higher product costs. In 2023, Tim Cook told The Wall Street Journal that price pressures make all supply options worth considering and said everything needs to be on the table. Apple previously sought clearance from the Trump administration to purchase chips from CXMT and YMTC, and in 2022 it evaluated YMTC NAND chips for iPhones sold in China, a move that drew warnings from US senators. The Pentagon added CXMT to its blacklist over alleged connections to the People’s Liberation Army, and the company continues to negotiate with US officials to broaden the use of its products. The outcome of the lobbying request remains unclear.

TikTok said it is testing improvements to its spam detection system to identify accounts that post AI‑generated content in categories such as politics, current events, financial advice and medical information. The platform is participating in the C2PA Steering Committee to support broader industry standards for content provenance. The enhancements aim to reduce the spread of AI‑generated material that could undermine public trust or well‑being. TikTok did not specify a launch date for the updates. The company noted that the changes are part of ongoing efforts to maintain platform integrity.

Front. Psychol., 18 June 2026 – A study published in Frontiers in Psychology on June 18, 2026 found that adolescents who perceive their caregivers as frequently distracted by devices score higher on the Device Attachment Interference Scale (DAIS) and report greater insecure attachment to those caregivers. The research addressed a gap in the literature on technoference and phubbing within parent‑child relationships, where prior work has focused on romantic and peer interactions. Investigators validated the DAIS with 600 U.S. adolescents aged 12‑17 recruited online in August 2025, confirming a unidimensional factor structure using exploratory and confirmatory factor analyses. Higher DAIS scores were associated with both anxious and avoidant attachment to mother‑like and father‑like figures after controlling for age and gender. The sample showed diverse demographic representation, with 75.5% identifying as White and other groups represented in smaller proportions. The study employed a cross‑sectional design, preventing causal inference, and relied on self‑report measures that may overlap with attachment constructs. Limitations include the single‑time‑point assessment, lack of longitudinal data, and insufficient sample size for measurement invariance testing across caregiver roles. The authors suggest future longitudinal and multi‑informant research to clarify directionality and examine whether reducing caregiver device use can improve adolescent attachment security.

CarPlay offers customization of displayed apps, wallpaper selection, connectivity options, and new interface features. To customize, users open the iPhone Settings app, navigate to General > CarPlay, select the vehicle, and choose Customize. They can hide apps by tapping the red minus sign and reorder icons by dragging. Wallpaper options are accessed through the CarPlay Settings app under Wallpaper, where Apple provides a limited set of pre‑installed images. CarPlay functions via wired or wireless connection; wired connections generally provide faster, more reliable performance, while wireless may introduce latency during media or Siri use. Some adapters, such as the CarlinKit 5.0, enable wireless functionality in vehicles lacking native support for a price under $60. Interface appearance can be set to Light, Dark, or Automatic mode in CarPlay Settings. Driving Focus mode, configured in the iPhone Settings, suppresses notifications and can be set to activate automatically when CarPlay is connected; users may allow specific contacts and enable an automatic reply. Announce Messages with Siri reads notifications aloud and allows voice replies through Apple’s Messages app and compatible third‑party apps. SharePlay functionality enables multiple users to join a session via a QR code displayed on the Now Playing screen, allowing contribution of songs and queue management regardless of individual music subscriptions. iOS 26 introduced widgets to CarPlay; users manage them through Settings > General > CarPlay > Widgets, where they can add, reorder, and toggle Smart Rotate and Widget Suggestions. Live Activities appear on the CarPlay screen by default but can be disabled in Settings. Pinned Messages can be turned off in CarPlay Settings, and icon appearance modes include Default, Dark, and Clear. The article concludes by asking readers to share their own tips.

Researchers at the University of Tübingen concluded that Homo floresiensis, the diminutive hominin nicknamed "Hobbits," likely scavenged carcasses of pygmy elephant relatives (Stegodon) left by Komodo dragons rather than hunting them themselves. The assessment is based on bone assemblages from Liang Bua cave on Flores Island, where both H. floresiensis fossils and Stegodon remains are found in the same sediment layers.

Analysis of the bones revealed two distinct types of marks: cut marks from stone tools appearing on less desirable meat sections, and shallower, wider tooth marks matching the feeding behavior of Komodo dragons, which grip prey and shake their heads to tear flesh. Experiments with a Komodo dragon at Zoo Atlanta produced similar bite patterns when feeding on a goat carcass, supporting the interpretation that the dragon's feeding left distinctive marks on the Stegodon bones.

The study found no evidence of fire use in the layers associated with H. floresiensis, indicating that any meat consumed was eaten raw. Researchers calculated that hunting large prey would have required disproportionate time and energy compared to the caloric return, making smaller prey such as giant rats a more efficient food source, which aligns with the abundant rat remains in the cave.

The findings challenge earlier assumptions that H. floresiensis were capable big‑game hunters and that they used fire, and they add nuance to the ongoing debate about the species' ancestry. While the Hobbits' skeletal features show some similarities to Homo erectus, other traits suggest a closer relationship to earlier hominins such as Homo habilis or Australopithecus. The study does not resolve whether H. floresiensis descended from Homo erectus or an earlier, less specialized hominin, and further evidence is needed to clarify its place in the human family tree and the origins of its stone tool tradition.

The 1X NEO home robot, a $20,000 device introduced in 2025, displayed a new pair of artificial hands on July 9, 2026, featuring 25 degrees of freedom in the fingers, palms and thumbs. A robot that can assist with household chores must be able to manipulate objects designed for human hands. Prior versions of the NEO used simpler grippers, limiting its ability to handle fine tasks. The new hands enable the robot to assemble Lego bricks and perform other precise manipulations, as shown in a video released by 1X. The increased dexterity could expand the robot’s usefulness in home environments, where varied object shapes are common. While the demonstration highlights technical progress, the robot’s price and limited market availability continue to restrict its adoption in consumer households.

Apple has stopped signing several older iOS versions for legacy iPhone and iPad models, effective July 9, 2026. The change prevents over‑the‑air updates and IPSW installations for the listed software releases. According to a post by Aaron Perris on X, Apple ceased signing iOS 7.1.2 for the iPhone 4 (CDMA) and discontinued OTA support for iOS 6.1.3 and iOS 8.4.1 on the iPhone 4S, as well as IPSW support for iOS 9.3.5 and iOS 9.3.6 on the same device. Additional versions no longer signed include iOS 8.4.1 for the GSM and CDMA iPhone 5, iOS 10.3.3 and iOS 10.3.4 for the iPhone 5 (GSM and CDMA), iOS 8.4.1 for the iPad mini Wi‑Fi + Cellular, iOS 8.4.1 for the iPad mini Wi‑Fi + Cellular MM, iOS 6.1.3 and iOS 8.4.1 for the CDMA iPad 2 (Wi‑Fi + 3G), and iOS 8.4.1 for the iPad (4th generation) Wi‑Fi + Cellular and Wi‑Fi + Cellular MM, with corresponding IPSW support for iOS 9.3.5 and iOS 9.3.6 on those models. Apple resumed signing the affected versions at 3:28 p.m. ET on July 9, allowing users to restore or downgrade to those iOS releases again. The restriction impacts a limited portion of the user base, primarily those maintaining legacy devices for testing, app compatibility, or preservation. The update reflects Apple’s routine practice of ending support for older software on devices that no longer receive security patches.

GLM 5.2, an open‑weights AI model, prepared a quarterly VAT return for a UK small‑to‑medium enterprise in 68 minutes, processing 59 transactions at a cost of $2.73. The return was off by 7 pence from the human‑prepared figure.

The benchmark used the first quarter of 2026 books for Vineyard Finance, a UK company that prepares its accounts internally. Transaction data and receipts were extracted by Claude Fable 5 and fed to GLM 5.2, which ran on a Google Cloud instance with internet access and a pre‑authenticated CLI tool. The model was evaluated on six criteria per transaction, including transaction type, chart‑of‑accounts category, VAT treatment, VAT amount, reverse‑charge VAT, and receipt attachment.

Across the quarter the model scored 354 checks. It failed 20 checks in 18 transactions. One serious error involved the classification of founder shares: the model recorded the payment as a capital account rather than as unpaid share capital, a distinction that could affect legal and audit outcomes. The amount involved was about 10,000 GBP.

Fourteen of the remaining errors resulted from confusing zero‑rated VAT with tax‑exempt VAT, a subtle distinction that does not affect the VAT liability but may impact compliance. Three additional issues concerned the handling of split transactions where a payment was divided across multiple currency balances; the model sometimes allocated VAT inconsistently, though the impact was minimal.

The model consistently performed well in other areas. It correctly classified most transactions to the appropriate account, never attached an incorrect receipt, and demonstrated the ability to resolve ambiguous inputs such as transfers between accounts or purchases disguised as card payments.

These results suggest that AI can now perform routine bookkeeping tasks with accuracy comparable to a skilled accountant, though further scaffolding is needed to integrate the technology into UK SME workflows. A beta version of the product is available at toot-books.com, and inquiries can be directed to adam@vineyard-finance.com.

Mitchell Hashimoto, known for creating Vagrant, Terraform and other HashiCorp tools, spoke in a July 2026 interview with Alex Alejandre about his new terminal emulator Ghostty and the Zig programming language.

He explained that after fifteen years of building command‑line applications he wanted to understand terminal internals and to work on GPU programming and desktop systems using Zig.

Ghostty is a cross‑platform terminal designed to be fast and feature‑rich while supporting composition of applications. The project began as a private beta and was shared through a Discord group before its public release. Hashimoto proposes an n‑screen API that would allow multiple terminal panes with independent grid sizes and a button protocol that can send commands even after scrolling. He notes that current terminal designs separate primary and secondary screens, which limits functionality, and that there is no standards body governing terminal protocols. He also mentions experiments with Wayland integration and the Nushell ecosystem, which he decided not to adopt.

He said he does not aim to push terminals to extreme capabilities and believes each platform should remain focused on its strengths. He argues that terminal applications should be easy to compose, have clear security models, and benefit from improved automation. He draws on concepts from several ecosystems when designing libraries, emphasizing the importance of “nouns” in API design.

Hashimoto said he is a primary user of Ghostty, which guides his decisions, and that open‑source projects carry no warranty or obligation to users; forks are encouraged if users need changes. He noted that he feels an obligation to fix problems but not to accommodate every request, and that his limited time, due to a newborn child, influences his development schedule.

He suggested that learning C can help developers understand computer fundamentals and that Zig may be a suitable next step for those seeking low‑level control.

Former OpenAI employee Daniel Kokotajlo warned in a 2025 podcast interview that a scenario he called “AI 2027” described a possible future in which the artificial intelligence competition between the United States and China could result in human extinction or the concentration of power in a single government. Vance, who said he had read the “AI 2027” paper, cited the scenario during the interview. In July 2026, Kokotajlo’s nonprofit, the AI Futures Project, released a new proposal named “AI 2040: Plan A,” which proposes an international agreement to delay the development of AI systems surpassing human capabilities until 2040. The proposal recommends that companies make their advanced models publicly accessible and allow external review, a requirement that may face resistance from industry. Kokotajlo said he anticipates opposition from firms and noted that mentions of “AI 2027” have increased as previous predictions have proven accurate. He added that a future U.S. president might be concerned about AI risks, and that “Plan A” could become part of the policy debate. The plan does not guarantee implementation, and its success would depend on cooperation between the United States and China.

AMD has introduced the Ryzen 7 4700LE, an eight‑core processor based on the Zen 2 architecture, for OEM use only.

The chip was listed on AMD's product pages on March 25, 2026, and replaces the integrated graphics found on earlier Ryzen 7 4700G models.

It features a boost clock of up to 4.2 GHz, a 65 W TDP, 12 MB of L2+L3 cache and a maximum temperature of 95 °C, and it does not include any onboard GPU. The processor is intended for system integrators and is not sold as a standalone component; a prebuilt gaming PC on Amazon currently offers the chip paired with an RTX 3050 for $799.

The launch reflects AMD's strategy of extending its existing product lines with older silicon, allowing OEMs to incorporate the chip into new systems without requiring additional development.

Cooler Master today announced the launch of the V8 ACE 3DHP single‑tower CPU air cooler, available immediately in separate Intel and AMD platform versions. The hybrid design combines two 3D heat‑pipe elements with four conventional heat pipes to move heat from the CPU to the fin stack without increasing the cooler’s physical size, and Cybenetics ranked it first among single‑tower air coolers in its database and second overall versus larger dual‑tower models. The unit uses dual 120 × 30 mm PWM LCP fans with Loop Dynamic Bearings and a tool‑free sliding fan mechanism that allows easier installation and better access to memory. Visually it features a non‑LED industrial style inspired by a V8 engine. The Intel version supports LGA1851, 1700, 1200 and 115x sockets, while the AMD version supports AM5 and AM4 sockets. Cooler Master offers a six‑year warranty, and the cooler is priced at $119.99 or €109.90. Cooler Master says the design provides a high‑performance single‑tower alternative that competes with larger dual‑tower coolers while maintaining broader case and RAM clearance.

The article provides an overview of Lisp, a programming language created in the late 1950s that uses a prefix notation with parentheses.

Lisp is one of the oldest high-level languages still in use and has influenced many subsequent language designs.

It employs symbolic expressions, or s-expressions, which are either atoms or lists, giving the language homoiconic properties where code is treated as data.

Learning to read and write Lisp requires understanding its syntax, packages, symbols, the REPL, conditional forms, and macro definitions.

Paul Graham introduced the term Blub paradox to describe why programmers of less expressive languages may find it difficult to appreciate Lisp's capabilities.

Macros allow developers to define new language constructs; for example, a macro can create a while loop by combining the existing loop macro, preserving the condition and body forms for delayed evaluation.

The macro expansion demonstrates that arguments are not evaluated until execution, unlike a function which evaluates arguments immediately.

The language’s live system, accessed through a Read‑Eval‑Print Loop, lets programmers redefine functions, variables, or macros at runtime without stopping the program for compilation.

This live environment supports immediate testing and debugging of code units.

Examples show how macros can be used to build domain-specific languages: a html macro enables users to write HTML fragments using standard Lisp forms, and a graph DSL allows users to describe mathematical curves and points with concise code.

Software such as AutoCAD's AutoLISP and the Emacs editor illustrate how Lisp's extensibility has been applied to automate tasks and create integrated environments.

The article notes that earlier expectations for Lisp to become a mainstream language have not been realized, yet it has remained in active use for more than six decades and continues to be valued for its design.

Its long-term survival highlights the enduring relevance of its features, and learning Lisp can broaden a programmer's understanding of language possibilities.

A scientist at MIT and an associate professor at the University of Illinois have created a model to verify compliance with the Outer Space Treaty’s prohibition on nuclear weapons in orbit. The treaty, established in 1966 and ratified by 117 nations, bans nuclear weapons in space, but verifying compliance is difficult. The model, published in Nature, simulates interactions between high‑energy protons trapped in Earth’s magnetic field and fissile material, producing detectable neutron signals from nuclear weapons. A CubeSat weighing up to 18 kg could theoretically detect these signals from a distance of 4 km after about a week of observation. The study notes that further engineering proof‑of‑concept work is required to confirm feasibility and that additional research is needed to achieve high technical readiness.

Apple’s iOS 27 beta includes code indicating that its Wallet car key feature may soon support vehicles from Xiaomi and Lucid Motors, two automakers not previously listed. The code was spotted by Aaron Perris for MacRumors in iOS 27 beta 3, released earlier this week. Apple first announced Lucid Motors would support car key at WWDC 2025, but no further updates were reported until now; Xiaomi had not been mentioned previously. No specific car models from either company have been identified, and the timeline for a public rollout is unclear, though the inclusion of the feature in the beta suggests a possible September launch with the final iOS 27 release. Earlier reports noted that Apple’s backend was updated to add Volkswagen to the car key list. The car key capability lets users add their vehicle’s key to Apple Wallet on iPhone and Apple Watch, enabling lock, unlock, and start functions through passive entry, proximity-based entry, or remote control via the Wallet app. Express Mode allows use without Face ID, Touch ID, or a passcode, and keys can be shared via Messages, Mail, and AirDrop. Apple has not confirmed which models will be supported or when the feature will become generally available.

A 9to5Mac poll of 6,400 readers asked whether a vehicle without Apple CarPlay would influence a purchase decision. The poll found that 11.42% of respondents (734 votes) said they would consider buying a car without CarPlay, while 79.94% (5,138 votes) said they would not, and 8.64% (555 votes) responded maybe.

CarPlay has been a standard feature in many new vehicles, and General Motors announced plans to discontinue its use. Apple introduced next‑generation CarPlay at WWDC 2021, stating that 79% of U.S. buyers consider CarPlay‑capable vehicles.

Comments on the poll indicated that many Tesla owners cited the brand's integrated software as sufficient, noting that Tesla’s own system includes an Apple Music app. Some respondents said other Tesla advantages outweighed the lack of CarPlay.

The poll reflects consumer preference for CarPlay, suggesting that its presence remains an important factor for most car buyers, though Tesla drivers appear less dependent on it.

Alan Bellows, founder of the technology blog Damn Interesting, announced a one-time GoFundMe campaign to raise enough money to return to full-time writing for the next year.

He previously supported the site with part-time engineering work while producing long-form articles and a podcast; in recent years, full-time engineering positions have become scarce, leading him to take a permanent full-time job.

The campaign aims to collect a sum equal to his previous part-time earnings, allowing him to devote more time to the blog and podcast; the fundraiser is separate from the site’s regular donation program, which covers hosting and operational costs.

The post also includes an image of a 20-sided die from a 1950 Magic 8 Ball novelty toy, illustrating the probability distribution of its yes, no, and unclear responses; Bellows noted that the toy is twice as likely to show a yes answer.

Bellows invited readers to contribute via the GoFundMe link or the site’s donation page if they wish to support the project.

Michigan health officials reported 1,251 cases of Cyclospora cayetanensis as of July 9, with 44 patients hospitalized, while Ohio health authorities reported more than 500 cases. The parasite, a microscopic protozoan transmitted through contaminated food or water, typically peaks in June and July and normally accounts for 2,000 to 5,000 U.S. cases annually. Michigan historically records about 50 cases per year, making the current tally more than twenty times the usual amount. The outbreak began with two cases on June 22 and rose to 572 cases by July 4, reaching a single‑day high of 239 on July 8. Cases are concentrated in southeastern Michigan, where multiple jurisdictions are interviewing patients to identify the source. Investigators have not determined how the infection is spreading, and cases continue to increase.

KDE Plasma 6.6.6 was released on July 9, 2026 as a bug‑fix update to the 6.6 series. The developers said the release contains numerous fixes and is likely the final update for the 6.6 series. Focus has shifted to Plasma 6.7, which was released in mid‑June and will be the last version to support X11 sessions. Plasma 6.8 is scheduled for mid‑October and will be Wayland‑only. Most default installations now use Wayland, though users can still install packages to enable X11. Wayland blocks the window manager from restoring window positions at login, according to developers. The XOrg server, maintained through the XLibre fork, has published 27 releases since its inception. The version number 6.6.6 references the biblical “Number of the Beast,” the developers said, adding that the naming was coincidental. Users of Plasma 6.6.5, including those on FreeBSD 15.1, can apply the update to receive the bug fixes.

Valve introduced the Steam Machine, a PC‑based console, at a price of $1,049 without a gamepad and $1,128 with one, aiming to bring PC gaming to the living room.

Game consoles have historically offered a simplified experience by running proprietary software, whereas the Steam Machine combines PC hardware with a console‑style interface.

The device runs SteamOS, a Linux‑based operating system, and can be operated with any standard gamepad. It is physically smaller than competing consoles and operates quietly. However, its price is roughly double that of a PlayStation 5, and performance benchmarks show no substantial advantage over a five‑year‑old PS5 when games are run at comparable settings. Users must manually configure games, install compatibility layers such as Proton, and manage driver updates. Reported issues include occasional reboots, failure to suspend sessions, incomplete game downloads, and HDMI compatibility problems. Valve has issued driver updates that improve ray‑tracing performance and plans to support FSR 4 upscaling. The system supports multiple monitors in desktop mode, external storage devices, and removable faceplates that can be 3D‑printed. Stock levels are constrained by memory supply shortages, with Valve securing about two‑thirds of its projected inventory.

Valve has indicated that ongoing updates will address current limitations, but the Steam Machine’s readiness for broad consumer adoption remains uncertain.

Accenture confirmed an isolated incident after a cybercriminal listed alleged internal data for sale on a cybercrime forum on July 6. The listing, titled “Accenture Data Breach,” was posted by a user using the handle “888” and claimed the archive contained 35 gigabytes of source code, cryptographic keys, Azure personal access tokens, storage access keys, configuration files and other technical material. The seller offered the data in exchange for Monero and included a screenshot of a cloned Azure DevOps repository named “121123_AtriasTalentAcademy” hosted on a censored accenture.com domain. Accenture spokesperson Andy Rowlands said the company had remediated the source of the issue, found no impact to operations or service delivery, and declined to provide further technical details. The Register asked how the compromise occurred, which systems were affected, and what data was taken; Accenture did not respond. SOCRadar noted that the same threat actor had previously claimed responsibility for a 2024 incident involving data linked to more than 32,000 current and former employees, though the legitimacy of that claim remains unverified. The scope of the alleged data set and whether it originates from a single repository or a broader exposure have not been clarified. Accenture has not announced additional findings, leaving the full extent of the incident unresolved.