Orson Amiri

  Rabbit is taking its original AI-agent vision beyond the R1 with a new software-focused approach. The company has launched Rabbit OS3, an agentic operating system designed to work across computers, the web, messaging platforms, connected devices, and the R1 itself. Instead of requiring users to perform tasks manually across different applications, the system is designed to understand an intended outcome and coordinate AI models, tools, files, skills, and connected machines to complete the work. Rabbit OS3 became generally available on September 22, 2026, following an invite-only beta, and users do not need an R1 to access it. At the center of the platform is a conversational AI agent that can work with connected computers through Rabbit Agent, a local program available for Windows, macOS, and Linux. Depending on the task and permissions, the agent can interact with files, terminal commands, desktop applications, and websites. Rabbit says one account can connect up to five devices...

  Apple is reportedly experimenting with a screenless fitness band that could bring it closer to WHOOP’s minimalist approach to health tracking. Reports suggest the prototype features a fabric strap and a sensor-equipped computing module designed to monitor metrics such as heart rate and other health data. However, Apple has not confirmed the product, its specifications, pricing, or branding, and reports indicate that a commercial launch may not happen before 2028. The project is still considered experimental, so it remains unclear whether Apple will eventually bring it to market. The idea of removing the screen may seem unusual for Apple, but WHOOP has already built its wearable strategy around exactly that concept. A screenless design can reduce distractions, support longer wear, and allow the hardware to focus primarily on sensors and health tracking. WHOOP 5.0 offers more than 14 days of battery life and focuses on continuous monitoring across sleep, recovery, strain, and other...

  Meta is reportedly testing a new AI feature called “super sensing” for its smart glasses that could make the devices far more aware of their surroundings. According to the Financial Times, the prototype could continuously capture audio and take photos every few seconds, allowing an AI assistant to help users remember things they saw or heard during the day. The reported feature could potentially answer questions such as where a user left their keys or what happened earlier, turning smart glasses into a more persistent memory assistant. However, Meta has not officially announced the feature or confirmed that it will reach consumers. The biggest concern is not what the glasses can remember, but what they could capture without other people knowing. Meta is reportedly considering keeping the recording indicator LED off while super sensing is active, which could make it difficult for bystanders to know when they are being recorded. Privacy concerns are already surrounding camera-equip...

  Choosing between a Gaming PC and a console comes down to how much control you want over your gaming experience. Consoles use fixed hardware, giving developers a predictable platform to optimize games for, while Gaming PCs allow players to customize components such as the CPU, GPU, RAM, storage, and cooling. That flexibility also brings more configuration and troubleshooting. In 2026, the price gap is less straightforward than before, with Nintendo Switch 2 priced at $449.99 in the U.S. through August before rising to $499.99 from September 1, while Xbox Series X and PS5 Pro are priced at $799.99 and $899.99, respectively. Performance is another major difference, but average FPS does not tell the entire story. Modern PC GPUs can deliver substantially higher frame rates, particularly at 1440p and high-refresh settings. For example, Tom's Hardware's June 2026 testing recorded averages of 96.4 FPS for the RTX 5070 and 104 FPS for the Radeon RX 9070 at 1440p Ultra without upscalin...

  Ox Alpha has quickly become one of the most discussed AI models among developers, despite arriving without a public developer name or traditional launch campaign. The model appeared on OpenRouter on August 20, 2026, through an unnamed third-party provider and is designed for coding, complex reasoning, sustained agentic tasks, and production workloads. Its biggest headline feature is a 1,048,576-token context window, along with support for up to 131,072 output tokens. It can process text, images, and video, while also supporting tool calling and structured JSON responses. The model's combination of capabilities and anonymous origins helped fuel its rapid popularity. OpenRouter currently lists prompt and completion tokens at zero cost, while OpenCode promoted free access for a limited period. OpenCode also reported significant usage, including 26 trillion Ox Alpha tokens processed in four days, highlighting the level of interest surrounding the model. However, the current free pric...

  A serious warning about advanced AI unexpectedly became one of X’s strangest viral memes. Former Anthropic researcher Jacob Coxon resigned after criticizing Anthropic and OpenAI over their approach to increasingly capable AI systems. He warned that companies were taking major risks while pursuing self-improving superintelligence. His resignation post quickly attracted widespread attention, reportedly passing 150 million views. What began as a serious AI safety statement soon became the foundation for a new internet copypasta. X users began copying Coxon’s dramatic resignation format and replacing the original AI catastrophe scenario with deliberately ridiculous threats. The structure remained largely the same: announce a departure, describe an alarming technological development, and warn about its consequences. The difference was that the supposed dangers ranged from Looney Tunes-style fake tunnels to self-closing accounting books, song lyrics, crypto trading, and fictional workp...

 Meta Muse Code and Anthropic’s Claude Code are taking AI-powered software development beyond simple code suggestions, giving developers agents that can plan, execute, test, and manage complex coding tasks. Meta introduced Muse Code alongside Muse Spark 1.2 in August 2026 and later upgraded it to Spark 1.3 in September. The coding agent is designed for large repositories and long-running development tasks, while Claude Code has evolved around delegation, automation, and developer-controlled workflows. The biggest difference lies in how the two systems handle AI agents. Muse Code uses a primary agent supported by specialized background agents that can remain active throughout a session. This persistent approach is designed to reduce repeated context gathering and developer intervention during lengthy tasks. It also uses a local event log that records model calls, tool activity, approvals, and edits, allowing work to resume after interruptions. Features such as /plan , /grill , and /...

IFA 2026 once again proved why it remains one of the world's biggest stages for consumer technology. Held in Berlin from September 4–8, the event showcased an exciting mix of AI-powered devices, smart home innovations, futuristic wearables, and next-generation computing concepts. Much like the gadgets seen in classic sci-fi movies and TV shows, this year's lineup introduced products that blur the line between imagination and reality. Among the biggest highlights were devices that rethink everyday technology in unexpected ways. Philips introduced the flexible Hue Liane 360° Rope Light that transforms lighting into a decorative design element, while Hypershell showcased the X Ultra S, an AI-powered consumer exoskeleton built to make hiking and outdoor adventures easier. Dyson also grabbed attention with its CameraJet toothbrush, which combines an integrated camera, AI, and liquid flossing to improve oral care. Lenovo demonstrated its innovative ThinkBook Project Swan, a concept l...

  Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1, marking another significant step in the evolution of its frontier AI models. While both models are built on the same underlying architecture, they are designed for different audiences through distinct safety controls. Claude Fable 5.1 is broadly available for developers, enterprises, and paid Claude users, whereas Claude Mythos 5.1 is reserved for vetted organizations working on advanced cybersecurity and life sciences projects. Beyond performance improvements, Anthropic has focused on making long-running AI workloads more efficient by significantly lowering prompt cache-read costs. The biggest upgrades in Claude Fable 5.1 revolve around extended coding sessions, autonomous agentic workflows, and complex knowledge tasks. The model can work across entire codebases, perform code reviews, generate and verify tests, identify root causes instead of temporary fixes, and interpret charts, tables, and diagrams within docume...

  Apple’s M6 chip represents a major step forward in Apple’s silicon roadmap, moving from the M5’s third-generation 3nm process to a new 2nm architecture. Introduced with the new Mac mini, the M6 comes with a 12-core CPU, 12-core GPU and a Dual 16-core Neural Engine, while the M5 features a 10-core CPU, 10-core GPU and a single 16-core Neural Engine. Apple claims the M6 can deliver up to 1.2x the multithreaded CPU performance of the M5, while its GPU provides nearly 30% more peak AI compute. Memory bandwidth also increases to as much as 170GB/s, although 16GB M6 Mac mini configurations continue to operate at 153GB/s. The M5 established a stronger foundation for on-device AI by adding Neural Accelerators to its GPU cores, alongside a 16-core Neural Engine and 153GB/s memory bandwidth. The M6 expands this AI-focused approach with additional processing resources and a second 16-core Neural Engine, which Apple says can provide up to twice the peak compute of the previous generation. Th...

  The World Humanoid Robot Games 2026 transformed Beijing's Ice Ribbon into a massive showcase for humanoid robotics, bringing together 2,056 robots from 666 teams representing 16 countries. Far more than a sporting event, the competition featured athletics, football, table tennis, gymnastics, dance, weightlifting, combat sports, and practical real-world challenges designed to measure how well robots could perform everyday tasks. Compared with the inaugural 2025 edition, the Games demonstrated remarkable growth in both participation and technological capability, highlighting how humanoid robotics is rapidly evolving beyond research labs. Beyond sporting contests, the Games focused heavily on real-world applications that tested robots in homes, hotels, restaurants, factories, retail stores, libraries, logistics centers, and emergency-response scenarios. Competitors were challenged to make beds, sort books, carry drinks, stock shelves, handle hazardous materials, extinguish fires, t...

  President Donald Trump has officially launched the process of creating the U.S. Space Academy by signing an executive order on August 28, 2026. Instead of establishing the academy immediately, the order forms a NASA-led Presidential Commission that will determine how the institution should operate, including its governance, curriculum, admissions process, and long-term implementation. The proposed academy is intended to prepare future astronauts, engineers, scientists, military personnel, and commercial space professionals for America's expanding role in space. Designed as a national center for technical education, leadership development, and public service, the academy aims to strengthen the country's future space workforce. The commission will also recommend accreditation standards, graduate service obligations, partnerships with existing educational institutions, and whether the academy should operate within NASA. Despite the announcement, there is currently no confirmed c...

  The rapid rise of artificial intelligence is transforming industries, but it is also placing unprecedented pressure on the global memory and storage supply chain. As AI companies and cloud providers expand their infrastructure, demand for high-bandwidth memory (HBM), DRAM, NAND flash, and enterprise storage has surged. To maximize profitability, manufacturers are allocating more production to AI-focused hardware, leaving fewer resources available for consumer devices. This shift is impacting far more than data centers. PC manufacturers, smartphone brands, automotive companies, gaming firms, and enterprise IT providers are all dealing with rising component costs and tighter supplies. Consumers are already feeling the effects through higher prices for RAM modules, SSDs, laptops, smartphones, USB drives, and other storage products. Since building new semiconductor fabrication facilities takes several years, increasing supply is not an immediate solution. The growing RAM And Storage ...