Apple announced new Mac mini and Mac Studio models with M6 and M5 chips on August 25, 2026. Pricing starts at $899 for the Mac mini and $2,499 for the Mac Studio. Pre‑orders are open and the products will be available on September 22.

The update follows the previous generation M4 models released in 2024 and 2025 and marks the first August product announcement since the 27‑inch iMac in 2020.

The Mac mini includes Wi‑Fi 7 and Bluetooth 6, 2.5Gb Ethernet as standard with an optional 10Gb upgrade, and supports genlock over USB‑C. The Mac Studio starts with 10Gb Ethernet. Both models feature an upgraded Neural Engine in the M6 chip that provides double the peak compute for AI workloads. The M6 Mac mini’s 12‑core GPU includes Neural Accelerators, delivering up to four times faster AI performance and twice the graphics speed compared with the M4 Mac mini. The M5 Ultra in the Mac Studio is the first Ultra chip with Neural Accelerators, offering 4.3 times the AI compute of the M3 Ultra. Storage performance on the Mac Studio improves with a PCIe Gen 6 SSD architecture, and Apple reports twice the SSD speed of previous models.

Upgrading a Mac mini to a 2TB SSD costs $1,000, exceeding its base price, while a 4TB SSD upgrade for the M5 Pro model adds $1,800. The M5 Pro Mac mini can be configured with an 18‑core CPU, 64GB RAM, an 8TB SSD and 10Gb Ethernet for $6,799. The M5 Ultra Mac Studio with a 36‑core CPU, 256GB RAM and a 16TB SSD is priced at $18,299. The M5 Ultra will support up to 512GB RAM, with the option arriving in late October; the cost for the 512GB configuration is expected to be at least $6,000.

Pre‑orders began on the announcement day and the devices will be available on September 22. Apple’s new Advanced Manufacturing Center in Houston, Texas, will begin producing the Mac mini this year, offering free training and educational sessions for small and medium businesses.

The announcements precede the anticipated iPhone 18 Pro event scheduled for September 9, and Apple has not indicated any changes to its leadership timeline.

OpenAI announced its first‑generation Jalapeno ASIC, which it says delivers 1.5 to 1.9 times more AI work per kilowatt than NVIDIA Blackwell chips and reduces latency compared with NVIDIA GB200/GB300 while challenging NVIDIA’s CUDA software ecosystem.

The chip, designed using OpenAI models and the Gluon programming language, is built on TSMC’s N3P process with an N3E I/O chiplet and HBM4 memory, providing 15.4 TB/s memory bandwidth per package and a systolic matrix engine that uses MXFP4 compression.

According to SemiAnalysis, the single reticle‑sized compute die achieves up to 13.4 PFLOPs of MXFP4 performance, compared with 17.5 PFLOPs for a similarly sized Rubin NVFP4 die. The design includes 2048 XPU scale‑up capability versus 72 for NVIDIA NVL72, and a rack‑scale configuration of 128 Jalapeno chips delivering 1.7 ExaFLOPs of 4‑bit processing and 27.5 TB of HBM4 memory per rack. Power consumption is rated at 700 W TDP but typically stays at or below 550 W during workloads. The architecture uses a single‑token prediction approach without speculative decoding and includes 64‑bit scalar cores with out‑of‑order pipelines and FP32/INT32 vector cores for high‑precision tasks.

OpenAI has not disclosed pricing or production timelines, and the performance advantage is based on tests run on its own model suite.

NVIDIA announced DLSS 4.5 Ray Reconstruction at Computex 2026, adding the 2nd generation transformer model to its ray tracing pipeline for GeForce RTX GPUs. The update replaces the previous in‑engine denoiser and is available on RTX 20, 30, 40 and 50 series cards. NVIDIA trained the model on expanded failure‑mode data using Blackwell GPUs, resulting in a 35 % increase in compute capability and 20 % more parameters while maintaining comparable performance. The new model provides deeper spatial awareness, more accurate lighting reconstruction, and improved temporal stability for ray‑traced and path‑traced content. Developers can fine‑tune temporal accumulation through the NVIDIA App, selecting a custom Ray Reconstruction preset. Testing with an RTX 5090 showed minimal frame‑rate variation, typically within plus or minus three frames per second across RTX 40 and older series. The technology is supported in titles such as Alan Wake 2, Cyberpunk 2077, Resident Evil Requiem, and others.

SpaceX announced plans to build a new spaceport at Starbase, Louisiana, with an investment of at least $100 billion and the creation of more than 3,000 jobs. The facility is intended to support thousands of Starship launches per year and will include propellant production, power generation, deep‑water shipping, vehicle processing and an airport. Construction has begun, according to SpaceX. The project will rely on Louisiana workers and local communities for labor. Environmental measures are part of the plan: SpaceX will work with state and federal agencies, landowners and conservation groups to protect shoreline, restore wetlands and preserve wildlife habitat. In Vermilion Parish, shoreline erosion rates range from 3.3 to 23 feet per year; the company will expand the state’s Coastal Master Plan and implement breakwaters to reduce wave energy. Plans call for thousands of acres of marsh creation using dredged material and offshore sediment, as well as interior marsh bank stabilization and canal reconstruction. Wildlife habitat on Pecan Island and surrounding wetlands, important for migratory birds, will be preserved, with monitoring and management coordinated with wildlife agencies and conservation organizations. SpaceX will not develop the entire footprint of the site and will maintain existing recreational activities such as hunting, fishing and birding. The spaceport is expected to increase the United States' capacity for multiplanetary launches and sustain national space priorities.

China has slowed shipments of germanium and quartz to Taiwan, creating supply constraints for semiconductor manufacturers and optical and robotics industries. The government imposed export controls on germanium in 2023 and on quartz in late 2024. Earlier, it restricted exports of gallium and germanium to the United States, lifted that ban, and then introduced a regime requiring exporters to disclose the customer and intended use of the material. Industry sources report that companies are having trouble obtaining germanium, quartz and permanent magnets because customs clearance times have lengthened. Some Taiwanese manufacturers have lost orders while waiting for shipments to be cleared. China's Ministry of Commerce has the authority to approve, reject or delay shipments after reviewing the disclosed customer information, effectively shaping the flow of these materials. The export restrictions remain in effect, and the resulting supply constraints are expected to continue as companies seek alternative sources.

Metro 2039, the next installment in 4A Games' post‑apocalyptic first‑person shooter series, was announced with a February 2027 release date and a new trailer at Gamescom Opening Night Live.

The Metro series, known for its atmospheric depiction of a post‑nuclear Moscow, follows a new protagonist called the Stranger who returns to the ruined city after years of exile.

Survivors have lived in Metro tunnels for 25 years, recently uniting under a leader named Hunter, who governs through propaganda and fear, while a dark enemy looms outside the tunnels.

4A Games said Metro 2039 aims to deliver a darker, more expansive experience, combining exploration, survival, stealth and combat across devastated wilderness and ruined districts.

The release window remains set for early 2027, with the game planned for multiple platforms.

Apple announced on August 25 2026 that a new Mac mini will be offered with the M6 and M5 Pro chips. Pre‑orders opened the same day, and the devices will ship on September 22.

The Mac mini line, first introduced in 2010, has been refreshed periodically with Apple‑designed silicon. The current generation uses the M4 chip, which succeeded the M1, M2 and M3 families.

The M6 model includes a 12‑core CPU with two additional performance cores, a 12‑core GPU with Neural Accelerators in each core, and a dual 16‑core Neural Engine, Apple said. It claims up to 4× faster AI performance, twice the storage and graphics speed, and 40 % faster CPU performance compared with the M4, according to Apple. Memory options begin at 16 GB unified memory and can be configured to 32 GB, with memory bandwidth up to 170 GB/s. The system also supports Wi‑Fi 7, Bluetooth 6, 2.5 Gb Ethernet with a 10 Gb option.

The M5 Pro model provides up to 18 CPU cores and 20 GPU cores, includes third‑generation ray tracing, and adds Neural Accelerators to each GPU core, Apple said. It supports up to 64 GB unified memory and 307 GB/s memory bandwidth. Apple reported up to 8.5× faster LLM prompt processing in LM Studio compared with the M2 Pro, up to 4× faster than the M4 Pro, and up to 4.5× faster rendering in Blender versus the M2 Pro, up to 1.4× faster than the M4 Pro. Image processing in Affinity shows up to 2.1× speed improvement over the M2 Pro and up to 1.5× over the M4 Pro.

The M6 model starts at $899 in the United States, with an education price of $799. The M5 Pro starts at $1,699, with an education price of $1,599. Pre‑order is available through apple.com/store and the Apple Store app in 30 countries and regions. Availability begins on September 22 at Apple Store locations and authorized resellers.

The new Mac mini adds higher performance for AI workloads and professional creative tasks while retaining a compact desktop form factor. Shipping will start on September 22.

Apple today announced a new Mac mini with M6 and M5 Pro chips, a refreshed Mac Studio offering M5 Max and M5 Ultra chips, a lower‑priced polishing cloth, and updated Magic Keyboards for Macs.

Apple described the M6 as its first 2‑nanometer processor and said the M5 Ultra is its most powerful chip to date.

Pre‑orders for the new Mac mini and Mac Studio are available immediately, with shipments and in‑store availability set for Tuesday, September 22. The polishing cloth and Magic Keyboard accessories are currently in stock.

Apple’s Houston Advanced Manufacturing Center, a 20,000‑square‑foot facility that produces AI servers and will begin manufacturing the Mac mini this year, opened earlier this month and provides free training and educational sessions for small and medium businesses.

The updates mark the first major refresh of the Mac mini line in almost two years.

X Corp. sent cease‑and‑desist letters to the open source project Nitter and its instances on August 24, 2026, demanding a permanent shutdown after alleging unlawful scraping of its data. Nitter is an open source tool that aggregates public X posts, removes advertisements and tracking elements, and lets readers view content without a X account. The service has been available since 2019 and has powered sites such as XCancel that provide direct access to X posts. In 2024 X imposed new API restrictions that temporarily took Nitter.net offline, after which the project restored its instances and resumed development. The cease‑and‑desist letter, which TechCrunch viewed, cites the Texas Harmful Access by Computer Act and the Lanham Act, alleging violation of X’s API terms and misuse of its data. X gave Nitter until 5 p.m. EST on August 25 to shut down. Nitter’s website, nitter.net, is currently offline and the project’s creator, identified as Zedeus, said he is seeking legal advice and will not comment on the specifics. He added that other Nitter instances have received similar letters. X’s action follows a pattern of larger social networks pursuing legal measures against third‑party readers. Meta has taken multiple scrapers to court, and many platforms now require users to log in through their official apps. The shutdown would remove a free, account‑free way for users to follow X posts, possibly leading them to create X accounts or stop using the service. The legal dispute remains ongoing, with no immediate resolution announced.

Google has added AI-generated summaries, called AI Overviews, to its search results. The feature appears after a query is entered, offering a concise answer while the traditional link list remains below.

Google Search remains the most visited website, handling billions of queries daily. Historically, the service displayed only a list of links, requiring users to click through to external pages for detailed information.

Liz Reid, Google’s vice president of search, said AI is driving the most significant upgrade of the Google Search experience ever. In May, Google announced that further AI capabilities would be rolled out. Users have reported varying levels of satisfaction with the summaries, and the company is collecting feedback to assess the impact on search quality.

Google plans to continue expanding AI features in search while monitoring user responses.

Ringg, a voice AI startup, raised $10 million from Peak XV Partners, extending its Series A to a total of $15.5 million. The company processes roughly 20 million call attempts each month and, according to a Truecaller study, more than 76% of Indian consumers prefer speaking with businesses over the phone.

Founded as a text‑to‑speech venture called DesiVocal, Ringg built its own speech models to offer enterprise voice agents. Its early customers included Cred, and it now serves startups such as Flipkart, Practo, Groww and PolicyBazaar.

The platform supports 1,200 clinics through Practo, enabling appointment booking, abandoned‑cart recovery and KYC checks. Voice calls still account for about 70% of Ringg's revenue, but the company is adding chat, WhatsApp and browser‑based support for clients like Shell. Co‑founder Siddharth Tripathi said the initial focus on low‑complexity tasks proved unsustainable, prompting a shift to higher‑value workflows. Peak XV principal Rishen Kapoor noted that Ringg’s in‑house model development allows end‑to‑end execution of complex enterprise processes.

Ringg plans to expand its AI stack, hire additional engineers and researchers, and deepen partnerships with global capability centers to sell automation capacity alongside human support.

Generalist, a robotics startup, announced a $200 million extension to its Series B, bringing total funding to $600 million and raising its valuation to $3 billion, according to two people familiar with the deal.

Founded in 2024 by former Google DeepMind researchers Pete Florence and Andy Zeng and former Boston Dynamics engineer Andrew Barry, the company has raised $600 million in total, including a $400 million Series B in June at a $2 billion valuation.

The round was led by 8VC and also included Radical Ventures, Nvidia, Union Square Ventures, Bezos Expeditions and AI researcher Fei-Fei Li.

Generalist is developing an AI foundation model that can work with various robots. Its Gen 1.5 model allows robots to learn new tasks from video demonstrations lasting three to twelve seconds.

The company is working with a handful of customers to tailor the model for specific use cases, according to a source.

Generalist competes with firms such as Physical Intelligence, valued at $11 billion, Skild AI at $14 billion, and Genesis AI, which is in talks to raise capital at a $3 billion valuation.

Investors see the technology as a potential “ChatGPT moment” for robotics, though some venture capitalists caution that a truly general robotics model may still be years away because robots cannot be trained on the full internet’s data.

OpenAI confirmed that Chris Malone, its former head of data centers, left the company last week, the Wall Street Journal reported.

Malone, who spent nearly five years at Meta and more than a decade at Google before joining OpenAI in March 2025, oversaw the execution of the company’s data center strategy.

His departure follows the launch of the Stargate Project, a $500 million U.S. data center initiative announced by the Trump administration and involving partners such as Oracle, Nvidia, SoftBank and Microsoft.

OpenAI said in a statement that it had recently reorganized its infrastructure organization to support the scale and pace of its work and that it maintains a strong, experienced data center team.

As part of the reorganization, Malone reported to OpenAI Vice President Sachin Katti rather than President Greg Brockman, according to the Journal.

Several other executives continue to lead data center efforts, including Uday Ruddarraju, head of the data center team; Brent Mayo, who directs the build and delivery program; and Spas Lazarov, a veteran of the data center and energy industries, who leads all data center engineering.

His exit adds to more than a dozen senior departures in 2026, a count that Business Insider placed at 13, with several occurring in the past month.

Other notable exits include the replacement of chief revenue officer Denise Dresser in February, the departure of Brad Lightcap, the former chief operating officer, in late February, and the resignation of Fidji Simo, the company’s product and business chief, in early March. Simo stepped down to recover from a chronic illness and remains in an advisory capacity.

The company also lost its head of ethics, Chloé Bakalar, in July, and announced the dissolution of its preparedness team last week.

Additional departures include former Sora lead Bill Peebles, who left after the project was shut down, and former chief marketing officer Kate Rouch, who resigned for health reasons.

Co‑founder Greg Brockman said the intense public scrutiny on the company makes every departure appear more significant than in other organizations.

OpenAI’s IPO, originally slated for 2025, has been pushed to 2027, and the firm is facing reputational scrutiny related to its valuation and profitability relative to large capital investments.

The ongoing turnover has not quieted concerns about the company’s leadership stability as it prepares for a public listing.

Warlock, a D&D‑based action adventure, debuted gameplay footage at Gamescom. The game was announced at The Game Awards last year and now presents eight minutes of in‑game video and new details about its design. Warlock is a single‑player third‑person title focused on spellcasting, rituals, exploration and puzzle solving. Players control Katri, a weapon master. She forms a pact with Tasha, an ancient witch. The pact grants access to abilities such as the Warlock’s Eye, which harvests raw magic from the environment. Raw magic can enhance existing powers or unlock new ones. Gameplay footage showed an open world with visual detail comparable to Baldur’s Gate 3, including cinematic sequences and dense environmental textures. Katri is accompanied by a devilish imp that follows her throughout the world. In the demo she visits a village, pays a beggar for information about suspicious activity in a nearby cellar, then transforms into a cloud‑like form to pass through a metal grate into an underground bunker. Her abilities let her manipulate the environment: she can collect broken stone fragments to create makeshift platforms, perform a super‑jump to reach higher areas, freeze enemies, and teleport behind them. Rituals can be conducted outdoors to control weather, for example calling rain to debuff fire‑based foes. Raw magic can also be harvested from larger enemies, such as a zombie ogre in the demo. While Katri wields a sword, the focus remains on magical abilities. The game includes a wide range of D&D spells and original creations, but it is not a CRPG. Combat is real‑time with no dice rolls. Spells such as Black Tentacles can solve environmental puzzles or be used in combat. Developers said the design encourages experimentation, with most abilities serving multiple purposes across traversal, combat and puzzle solving. In a Q&A session they described the game as an “exploration‑driven open world,” noting that the map is dense rather than impossibly large and that the team sought to avoid excessive handholding. Players can explore non‑linear paths, discover mysteries that function as side quests, and interact with NPCs to gain new leads, such as the beggar directing Katri to the cellar. Cantrips are available for free use, while more advanced spells require recharging. The setting is the Domains of Dread, a darker region of the D&D universe, and Katri is an outsider who relies on her imp companion and telepathic communication with Tasha, whose true intentions remain unclear. As a Warlock and weapon master, she can conjure various melee weapons, and armor sets found in the world may possess magical properties that alter abilities. The developers emphasized immersion and confirmed that the game will include an endgame allowing continued exploration and side content after the main quest. Enemy difficulty scales to indicate when players are not ready for certain zones, and progression aims to balance exploration with main‑quest advancement. An exact release date has not been announced, but the game is planned for 2027. KitGuru said the first gameplay showing is brief but promising, highlighting strong production values, versatile spell mechanics and a dense open world. Further details are expected as development continues toward its planned 2027 release.

Chris Malone, who led OpenAI’s data center initiatives, departed the company last week, according to reports from The Wall Street Journal and Bloomberg. Malone joined OpenAI in early 2025 after serving as a senior executive at Meta and Google, initially reporting to president and cofounder Greg Brockman. The leadership change follows broader reshuffling within the organization earlier this year. OpenAI has not issued a public statement on the departure. The report does not specify the reasons for Malone’s exit.

The 2nd World Humanoid Robot Games were held in Beijing from August 22 to 26 2026, featuring competitions in which humanoid robots sprinted 100 meters and performed high jumps that exceeded human world records, as well as events testing domestic and industrial tasks such as laundry, package delivery, hotel room preparation, library shelving, and firefighting.

The competition, organized by Chinese state media and the Beijing city government, follows the inaugural event held in 2025 and aims to demonstrate commercial potential of humanoid robots amid rapid development in China’s robotics sector.

In the 100‑meter sprint, several robots posted times faster than the human record set by Usain Bolt, though most could not decelerate after crossing the finish line and some collided with barriers or sustained fire incidents. A small‑group 400‑meter race was won by a Beijing‑based X‑Humanoid robot that employed a distinct posture developed through simulated reinforcement‑learning trials.

Additional challenges included washing and hanging laundry, moving wheeled luggage to hotel rooms, cleaning beds, restocking supplies, and collecting books for shelving, where robots generally operated more slowly than human workers. An outdoor firefighting exercise required teams to identify hazardous substances, shut off valves, locate a fire and extinguish it within 30 minutes; only three of twelve teams completed all steps.

Dipam Patel, a PhD candidate at Purdue University and researcher at the U.S. Army DevCom lab, said the performances indicate improved whole‑body control, noting that the robots must coordinate joints, hands, legs and torso to achieve the tasks. He added that many competitors remain teleoperated by humans and that autonomous systems still lack the ability to handle diverse or unforeseen edge cases.

Investment in humanoid robot companies totaled more than $6 billion in 2025, with Chinese firms leading deployment efforts while U.S. companies such as Boston Dynamics and Agility Robotics expand commercial applications in factories and warehouses. Analysts suggest that rigorous, repeatable testing—such as multiple runs of the same race—could better assess reliability, and that broader autonomy will require extensive data collection and algorithmic advances.

The event underscores that while humanoid robots can achieve specific performance milestones, substantial technical hurdles remain before they can operate reliably across varied real‑world environments.

Nokia 300 Charge, a feature phone announced on August 25, 2026, offers up to 40 days of standby power and can charge other devices, according to Nokia. Feature phones have historically provided extended battery life because of limited hardware demands, and the 300 Charge builds on that premise. The handset includes a built‑in flashlight on its upper edge and runs a simplified operating system that supports basic calls and texts. GSMArena reported the standby duration based on manufacturer specifications. The device targets consumers seeking long battery life without smartphone complexity, and its market performance remains to be seen.

Instagram announced on Tuesday that it is rolling out First Draft, a tool that automatically trims selected video clips and removes pauses to create an initial Reels edit.

The company said the feature can generate a draft in under 10 seconds, aiming to reduce editing time for creators and lower the barrier for those with limited experience.

First Draft is available initially on iPhone and can be accessed from the camera while recording or from the Reels gallery after selecting multiple clips. The automated process pulls together the best moments without manual trimming, allowing creators to refine the draft afterward. Similar automated editing functions already exist in Adobe Firefly's Quick Cuts and CapCut's Auto Cut, but Instagram's integration eliminates the need to transfer footage to a separate application. The announcement follows other creator tools released this year, including a Replace Audio function for feed posts, an AI assistant in the video-editing app Edits, and a Series feature that groups related Reels for sequential viewing. Instagram's head Adam Mosseri is scheduled to testify in a pending antitrust trial involving Meta, though the new feature is not directly tied to the legal matter.

The rollout marks another step in Instagram's effort to streamline short-form video creation as the platform prepares additional AI-driven tools for its editing suite.

Researchers at the Indianapolis Museum of Art and The College of Wooster have investigated the longevity of glow‑in‑the‑dark phosphorescent pigments used in art and fashion, focusing on the museum’s Stephen Sprouse collection.

Previous work by the same team examined daylight fluorescent pigments and found that optical brighteners degrade faster than the pigment dyes, causing colors to appear darker even when the dyes remain stable. That study highlighted the need to understand the distinct photochemistry of fluorescent materials.

The recent study shows that phosphorescent pigments consist of mineral‑based and inorganic compounds, differing from the organic dyes typical of fluorescent colors. Humidity has been identified as a major factor in their breakdown, often surpassing the effects of prolonged light exposure, which suggests adjustments to storage conditions may be necessary.

Sobeck said the effect is integral to the artists' visual intent. She and conservation scientist Gregory Smith have collaborated on the research for ten years.

The next phase will assess how long phosphorescent powders containing the white pigment lithopone retain glow after each charge, aiming to develop preservation guidelines for such works.

Dolly Parton died at age 80 on Tuesday after a brief battle with cancer, according to a statement from her nephew Brian Seaver posted on Instagram. She was pronounced dead at Vanderbilt-Ingram Cancer Center surrounded by loved ones.

Born in a one-room cabin in Tennessee, she was the fourth of 12 children in a family where her mother had 12 children by age 35 and her father worked as a farmer and construction worker. She received her first guitar at age eight from an uncle, began singing in church, appeared on local television at age ten, and cut her first record at age 13. After graduating high school in 1964, she moved to Nashville, initially working as a songwriter before being signed as an artist.

She achieved her first country No. 1 in 1971 with "Joshua" and reached broader success with "Jolene" in 1973 and "I Will Always Love You" in 1974. She amassed 25 US country No. 1 singles and 47 albums in the country Top 10, and a US No. 1 hit with "9 to 5" in 1980. Her catalogue includes the songs "Jolene" and "I Will Always Love You," which have been covered by numerous artists, including Whitney Houston, whose 1992 recording remains the best‑selling single by a woman with an estimated 20 million sales. She earned 10 Grammy Awards and received multiple nominations, including an Oscar nomination for "9 to 5" in 1980.

Parton founded the production company Sandollar in the 1970s, which produced television series such as "Buffy the Vampire Slayer" and the "Father of the Bride" film series. She invested in the Silver Dollar City theme park in Tennessee in 1986, renamed it Dollywood, and expanded the property with a spa, cabins, theatre‑restaurant and waterpark over the following two decades. The Dollywood Foundation, established in 1988, launched the Imagination Library in 1995, sending a monthly book to every child in Sevier County, Tennessee, and expanding nationwide in 2000 and internationally in 2006; the program distributed nearly 25 million books annually by 2022. That year, Jeff Bezos donated $100 million to the foundation.

She contributed $500,000 to the opening of a Sevier County hospital in 2006, supported relief efforts after the 2016 Smoky Mountains wildfires, and donated $1 million to Vanderbilt University in 2020 for Covid‑19 research that helped develop the Moderna vaccine. Parton also supported Black Lives Matter and transgender rights, opposing a North Carolina bathroom bill in 2016. In May 2026, she cancelled her Las Vegas residency due to health issues, and a Broadway musical based on her life was slated to open later that year after a run in Nashville.

In a recent tutorial, the author explains how to install OpenBSD on a DigitalOcean Droplet for approximately $4 per month. First, the miniroot image is downloaded from the OpenBSD website and its SHA‑256 checksum is verified. The image is then uploaded to DigitalOcean as a custom image, named, and the associated storage fee is noted. A Droplet is created with 512 MB memory, one virtual CPU, 500 GB of transfer and 10 GB of disk space, using the custom image as the boot source. An SSH key is added during creation, although DigitalOcean does not enforce its use. In the web console, the author boots the miniroot image, selects the default network interface, enables IPv4 and IPv6 autoconfiguration, sets a hostname, creates a root password, and disables root SSH login. A non‑root user account is created with a password, and the appropriate time zone and disk device (sd0) are selected; full‑disk encryption can be enabled, which requires entering a passphrase via the web console at each reboot. The installation proceeds by choosing a base set of packages—bsd, base79.tgz and man79.tgz—while removing X server, games and compiler sets. After the sets are installed, the system reboots, prompting for the encryption passphrase if enabled, and the author recommends using SSH with the non‑root account and disabling password authentication. The tutorial concludes with links to the OpenBSD Handbook and FAQ, and a note that the server runs on DigitalOcean for $4.24 monthly.

LatticeDB is an open‑source, single‑file graph database that supports native vector similarity search and full‑text indexing within a single query interface. The project, hosted on GitHub, provides embedded operation with zero configuration and a single‑writer model, allowing local applications to store and query graph‑structured, semantic, and textual data on a single machine. The database stores the entire dataset in one file and uses a write‑ahead log for durability. It offers a unified query language that can combine graph traversal, HNSW‑based vector search, and BM25‑ranked text search. Benchmark results on an Apple M1 processor show node lookups at 0.13 μs, 10‑NN vector search at 0.83 ms for one million 128‑dimensional vectors with 100 % recall, and full‑text search latency of 19 μs. Graph traversal benchmarks demonstrate speedups of up to 52× over SQLite for variable‑length paths. The project includes command‑line tools and bindings for Python, TypeScript, and Go. Use cases cited include local knowledge bases, agent memory, and retrieval‑augmented generation pipelines. LatticeDB’s single‑writer architecture limits concurrent writes from multiple processes, and it lacks full Cypher support and a mature ecosystem compared with established graph databases such as Neo4j. The license is MIT.

Newegg lists a bundle that includes a Ryzen 7 7700X3D processor, an Asus TUF B650E‑Wifi ATX motherboard, a single 16 GB G.Skill Ripjaws M5 Neo DDR5‑6000 stick, and a free Cooler Master Elite Liquid 240 AIO cooler for $568.99, a discount of $200 from the summed retail prices.

Single‑stick DDR5 RAM is listed at $70, and a promotional code can reduce the cost by an additional $20, resulting in a 32 GB dual‑channel kit priced around $320.

The Ryzen 7 7700X3D, an 8‑core 16‑thread Zen 4 CPU, is positioned among the top gaming processors, surpassing the Ryzen 7 5800X3D and the Intel Core Ultra 7 270K Plus while lagging behind the Ryzen 9 9800X3D and the Ryzen 7 7800X3D.

Tests indicate that a single‑DIMM configuration trails dual‑channel operation by approximately 3 % in gaming when used with an X3D processor, and outperforms DDR4 dual‑channel kits in single‑channel workloads.

The Asus TUF B650E‑Wifi motherboard includes power delivery suited to the 7700X3D, three M.2 slots (one PCIe 5.0), four SATA ports, integrated Wi‑Fi 6E, 2.5 GbE networking, and support for future AM5 processors.

Because individual component prices remain high, bundled offers such as this one provide a cost‑effective entry point to the AM5 platform.

The promotion is available for a limited period on Newegg.

Joe Shields, a staff writer at Tom’s Hardware, writes about motherboards and PC components.

A Raspberry Pi 5 running a 35‑billion‑parameter Qwen model has been deployed in a vehicle to provide an offline, locally hosted AI assistant that interacts with a GroupMind chat room and delivers vehicle‑related information.

The project, called CarWatch, was presented on Hacker News and details the hardware, software stack, and capabilities of the system.

The Pi 5, equipped with 16 GB of RAM and active cooling, runs the Qwen3.6‑35B‑A3B model quantized to 14.3 GB, delivering about 3.5 tokens per second for generation and over 25 tokens per second for prompts. The system operates without internet connectivity; the model, a 745‑page vehicle owner’s manual, and supporting files reside on an SD card, enabling retrieval‑augmented generation with page citations. A USB microphone captures voice input through whisper.cpp, which is processed locally and sent to the @gle agent in GroupMind rooms. The agent can reply to messages about departures, arrivals, trip summaries, and dashcam clips, using approvals from a phone or watch via the CodeWatch dashboard. Systemd services start the model server, voice listener, and dashboard automatically at boot, and the system updates itself hourly from the project repository or on demand through a dashboard button. Connectivity is maintained via a dial‑out tunnel for reachability behind NAT, with fallback to a Wi‑Fi hotspot or the car’s own access point. OBD data reading is implemented through a DoIP/ENET cable; the connection has been tested against a simulated gateway but not yet verified on a real vehicle. Dashcam clip retrieval is under investigation, while the dashboard UI and mirror icon strip are planned. The total hardware cost is approximately 300 €.

The build demonstrates a fully offline automotive AI solution, with many components proven in testing and others slated for future development.

Apple has reversed its plan to move Hide My Email addresses from @icloud.com to @private.icloud.com, keeping the addresses on the original domain. The feature, available to paying iCloud+ subscribers, generates disposable email addresses that mask a user’s real address while allowing messages to reach the inbox. Apple said law enforcement can still request the underlying address when needed. The reversal was announced on Monday after users criticized the change on social media and Reddit, saying it would undermine privacy. Daring Fireball’s John Gruber reported that Apple employees had opposed the change from the start. Earlier in 2025, a bug exposed users’ real email addresses, and Apple fixed the issue more than a year after discovering it. The update restores the original functionality and maintains the ability to comply with legal requests.

Apple announced the new Mac Studio with M5 Max and M5 Ultra chips on August 25, 2026, and began taking pre‑orders the same day, with shipments scheduled for September 22. The Mac Studio joins Apple’s lineup of desktop computers that use Apple‑designed silicon, following the M3 Ultra model introduced in 2025. The M5 Max provides an 18‑core CPU, up to a 40‑core GPU with Neural Accelerators, and supports up to 128 GB of unified memory with a memory bandwidth of 614 GB/s. Apple states the chip delivers up to 4.3× faster AI performance, 2× faster storage, 1.8× faster graphics and 1.3× faster CPU speed compared with the previous generation. The M5 Ultra adds a 36‑core CPU, an 80‑core GPU, and up to 512 GB of unified memory, offering up to 1.3× higher multithreaded performance than the M3 Ultra and up to 4.3× higher peak AI compute. It also supports up to eight external displays, Thunderbolt 5 ports with 120 Gb/s bandwidth, Wi‑Fi 7, Bluetooth 6, and a next‑generation SSD that provides up to twice the read/write speed of earlier models. The system runs macOS 27, which includes Apple Intelligence features such as Siri AI and a new Core AI framework for building and deploying models on Apple silicon. Developers can use MLX, an open‑source machine‑learning framework, and Xcode for end‑to‑end AI development. Pricing starts at $2,499 for the M5 Max configuration and $5,499 for the M5 Ultra, with education pricing available. Pre‑orders are open on apple.com and the Apple Store app in 30 countries, and the devices are expected to ship on September 22. A 512 GB memory option will become available in late October. Apple Upgrade offers monthly lease terms of $48.99 for the M5 Max and $110.10 for the M5 Ultra, subject to eligibility. AppleCare+ and AppleCare One provide coverage for accidents, battery replacement and priority support. Trade‑in and financing options are also available. The new Mac Studio expands Apple’s high‑performance desktop offerings and adds capabilities for AI research, video production and other professional workloads.

Pie, the social app founded by Bonobos’ creator Andy Dunn, announced a new Communities feature that lets users create permanent group spaces, shifting the platform’s focus from events to ongoing social interaction. The app, which began as an events service in Chicago and now has about 300,000 users, is backed by investors including Forerunner, Lightspeed, Accel and Ev Williams, the co‑founder of Twitter. The Community Homes feature provides group homepages with a community feed, admin and membership controls, a shareable homepage and event‑planning tools; any member may post events to the feed. The update follows earlier changes such as a feed that highlights attendees and a favorites feature that increased average time spent on the app from two to ten minutes. Dunn said the platform is meant to support real‑world connections similar to Reddit but without anonymity, and he highlighted a growing Gen Z interest in in‑person social clubs. Nadya Okamoto, former founder of the Zoomies run club, joined as co‑founder and chief marketing officer. The company plans to add group chats and a feature allowing casual posts about potential meetups. Pie is currently more active in markets such as Chicago, New York and Los Angeles.

Researchers at Cato Networks identified a campaign in which malicious actors placed sponsored Google ads linking to fake OpenAI Codex download pages, prompting Mac developers to run a command that installs malware. The ads directed users to a site hosted on Google Sites that displayed OpenAI branding but offered no legitimate installer. Instead, the page instructed users to open Terminal and paste a supplied command, which appeared to be an npm installation instruction but actually executed a multi‑stage infection.

The command decodes a Base64‑encoded URL, retrieves a shell script from an attacker‑controlled server, and pipes the script into zsh. The script downloads additional payloads, contacts the attacker’s server to report successful infection, and writes a universal Mach‑O binary to /tmp/helper. The binary disables macOS security flags that would normally flag suspicious downloads, allowing the malware to execute without warning. Cato noted that the final binaries are universal Mach‑O files, capable of running on both Intel and Apple Silicon Macs.

Cato found substantial similarity between this activity and Atomic macOS Stealer (AMOS), an infostealer previously spread through fake software downloads and malicious advertising. While Cato has not officially labeled the new malware as AMOS, it said the staging process and payload construction share fingerprints with that campaign. The attackers also used an iframe to pull the malicious content from their own infrastructure, checking the visitor’s operating system and referral path to serve harmless pages to users who do not match the intended profile. The decoy site offered both macOS and Linux download buttons, though observations indicated the malware chain was delivered only to Mac users.

Cato also discovered a related campaign that masqueraded as Anthropic’s Claude Code, sharing the same infrastructure. The ClickFix technique, which relies on convincing victims to execute malicious commands themselves, is increasingly used to bypass traditional attachment‑based defenses.

Cato Networks has not yet attributed the campaign to a specific group, and the full scope of the infection remains unclear.

TechCrunch said the lowest‑price registration window for TechCrunch Disrupt 2026 will close at 11:59 p.m. PT on August 25, ending a discount that can reach $300 per pass and offering larger group savings. The conference is scheduled for October 13‑15 at Moscone West in San Francisco and will host more than 10,000 founders, investors and other startup participants. Programming will focus on AI and include stages such as the Real World AI Stage, which examines robotics, autonomous systems, manufacturing, healthcare and defense applications, the Smart Money Stage covering fintech, stablecoins and AI‑driven financial services, and the Smart Systems Stage highlighting chips, compute, energy, networking and data‑center innovations. Returning stages include Disrupt, AI and Builders, with speakers such as Rivian CEO RJ Scaringe, Amazon SVP Panos Panay, Bluesky CEO Toni Schneider and COO Rose Wang. Attendees can access AI‑powered networking through the event app, deep‑dive side events, and the exhibit hall showcasing emerging technologies. Registration after the August 25 deadline will be subject to higher rates.

Waymo announced Tuesday it will begin testing its autonomous vehicles in Munich in the coming weeks, with a view to launching a commercial ride-hailing service in Germany by the end of 2027. The Alphabet-owned self-driving company plans to conduct manual driving and specialist-supervised tests to map local roads and adapt its system to German traffic conditions while working with local, state and federal officials to secure regulatory approvals. Waymo said it will invest in local fleet operations and create high-skilled jobs as part of its expansion. Uber and Pony.ai previously launched autonomous rides in Zagreb, Croatia, the first European city where self-driving vehicles could be booked through a ride-hailing app. Waymo raised $16 billion earlier this year at a $126 billion valuation as it expands its robotaxi business. The testing program is expected to inform regulatory discussions and support the development of a paid service in Germany. The rollout timeline remains subject to regulatory approval and the pace of technology adaptation, and Waymo has not set a definitive date for commercial service launch.

Amazon Prime Video has ordered a new RoboCop series starring Dan Stevens, which will consist of eight episodes. No release date has been announced. According to Deadline, Stevens will portray the lead character. The series is part of Prime Video’s expanding lineup of scripted content.

The article explains a tooltip interaction pattern designed to improve user experience on a web page that displays company logos next to interview questions. The author noted that tooltips opened instantly when the cursor moved, which felt abrupt. Adding a 200‑millisecond delay reduced the immediate appearances but caused a pause when switching between nearby logos. The proposed solution uses a warm state that enables instant tooltips after the initial delay, and a cold state that resets the delay when the page is idle. On mouse enter, a timer is started; if the warm state is active, the tooltip appears without waiting. On mouse leave, the tooltip closes and a 300‑millisecond cooldown timer begins. While the cooldown runs, subsequent hovers trigger instant tooltip display. The provider component tracks the warm flag with a reference, clears pending timers in markOpened, and starts the cooldown timer in markClosed. CSS rules skip animation when the instant flag is set. Code snippets in JavaScript demonstrate the onMouseEnter handler, the show function, the onMouseLeave handler, and the provider implementation. The author mentions using Radix and Motion libraries and provides a React example and a Claude skill for auditing similar codebases. The post frames the pattern as a small interaction improvement rather than a complete tooltip framework.