Recursive Self-Improvement or RSI is widely considered the holy grail of AI, and a wild rumor now suggests that Google's DeepMind - the unlikeliest of AI labs in the current paradigm - has achieved this feat, which then puts a cynical spin on a somewhat coordinated push from Anthropic, OpenAI, and Elon Musk to try to achieve regulatory capture of the broader AI sphere by advocating for pacing progress on AI models. Cynisim abounds as Anthropic, OpenAI, and Elon Musk all call for pacing the progress on AI, which critics argue only serves as a front for disadvantaging other AI […]

Read full article at https://wccftech.com/a-wild-rumor-says-google-is-on-the-verge-of-achieving-rsi-as-anthropic-openai-and-elon-musk-all-call-for-pacing-ai-in-what-might-be-a-cynical-attempt-at-regulatory-capture/

An aging Pixel smartphone will always have multiple use cases, and while Google’s Tensor might be the most underpowered SoC after Huawei’s Kirin series, it’s still capable enough of running smaller AI models if configured properly. For instance, one enthusiast made a cyberdeck housing a Pixel 7 and ran Qwen2.5-abliterate:3b, a 3-billion-parameter uncensored model, which has its safety filters completely removed. Unfortunately, even running a smaller model on a Pixel 7 requires patience, as the handset only manages 5 tokens per second Using Claude Code, Redditor “Levfo” generated the 3D-printed CAD files for the cyberdeck case used for the Pixel […]

Read full article at https://wccftech.com/older-google-pixel-smartphones-run-uncensored-on-device-ai-model/

The use of AI has enabled users to obtain feats that were never thought possible, such as vibe-coding a game featuring 27,000 lines of code, or enabling software compatibility between macOS and an HP printer without official software. New possibilities are being stumbled upon daily. For Claude Fable 5’s newest accomplishment, a user designed and manufactured an entire working PCB from scratch, but the process was far from perfect. To achieve a PCB design from start to finish costs $450 in API credits alone, making it a fun, but pricey project After going nowhere with advanced AI models like Opus 4.5 […]

Read full article at https://wccftech.com/claude-fable-5-ai-custom-pcb-design-demonstration/

AI model routers are emerging as one of the biggest threat vectors for data security, as evidenced by a shocking anecdote from a researcher, who was able to uncover a veritable treasure trove from a data dump on routing sessions related to Anthropic's Claude, with grave ramifications for vulnerable Chinese government entities and companies such as Huawei and Xiaomi. According to a researcher, just 6TB of data dump gleaned from routing sessions related to Anthropic's Claude can yield enough valuable credentials to hack top Chinese government agencies and companies For the benefit of those who might not be aware, LLM […]

Read full article at https://wccftech.com/a-researcher-buys-6tb-of-anthropic-claude-data-dump-from-a-china-based-llm-router-finds-enough-ammo-to-hack-xiaomi-huawei-and-chinese-government-agencies/

DeepSeek's engineers are a marvel, excelling in extracting every ounce of efficiency from the architectural constraints that characterize contemporary LLMs, as they appear to have done with the just-released V4.1 Flash, which is phenomenally competitive with the likes of OpenAI's GPT-5.6 Sol and Anthropic's Claude Opus 5 on coding and cybersecurity, while entailing just a fraction of their costs! The novel architecture of DeepSeek's V4.1 Flash model DeepSeek's V4.1 Flash model sports a fairly novel architecture. Given the inherent complexity, we'll try to explain the model's main architectural elements using easy-to-understand analogies. The V4.1 Flash is a multimodal Mixture of […]

Read full article at https://wccftech.com/deepseek-v4-1-flash-beats-openais-gpt-5-6-sol-and-anthropics-opus-5-on-coding-and-cybersecurity-at-an-86x-lower-cost-while-reducing-hbm-requirements-by-3-8x-and-ssd-ones-by-8x/

Running a standard 4-bit quantized version of Qwen3.8-27B requires you to have between 15GB and 17GB of available VRAM and system RAM combined, so when an intrepid user tried running the same AI model on his Windows laptop featuring just 12GB of memory, the immediate reaction would be that this is an impossible task. However, by running open-source software, the user managed to pool four devices to combine the total system RAM, which allowed Qwen3.8-27B to fire up successfully. The obvious trade-off is immensely slow token speeds of 2 toks/second, making the 2-bit quantized version of Qwen3.8-27B a sensible approach […]

Read full article at https://wccftech.com/qwen38-27b-ai-model-runs-on-12gb-laptop-by-pooling-memory-across-four-devices/

The acceleration within the AI space has sparked a revolution in the PC segment. We have seen major chipmakers embark on the "AI PC" journey, which involves dedicated client designs that mix Windows applications with dedicated and local AI subsystems, and then we have dedicated boxes for AI, which are made primarily for local acceleration of models and agents. Last year, NVIDIA introduced its DGX Spark PC, the first PC of its kind to tackle LLMs and Agents in what is dubbed to be an "AI Supercomputer". Soon after the launch, other manufacturers started to label their own Mini PCs […]

Read full article at https://wccftech.com/review/nvidia-dgx-spark-ai-supercomputer-impressions-the-future-of-ai-computing-in-tiny-boxes/

The World Health Organization (WHO) considers antibiotic and antimicrobial resistance (AMR) as one of the top health concerns of the 21st century, and Anthropic might have just discovered a viable vector to combat this emergent threat, at least as per a credible litany of rumor. Anthropic reportedly plans to announce the discovery of this new class of drugs just ahead of its IPO Most prevalent bacterial strains are rapidly evolving to develop resistance against run-of-the-mill antibiotics. This is partly a function of overuse and the innate ability of organisms to evolve in harsh conditions. Even so, a number of novel […]

Read full article at https://wccftech.com/anthropic-reportedly-discovers-new-antibiotics-that-are-effective-against-resistant-bacteria-and-timed-to-land-just-before-its-ipo/

Imagine you can tap into a government-run AI model for all of your requests, and that too ad infinitum. That was the stated purpose behind the "fanzha-ai-proxy" project that just got booted off GitHub, with many Chinese developers comparing it to "picking fleas off a tiger's head" in what is a nod to the severe penalties that China typically hands for such stunts. The premise behind a Chinese developer's fanzha-ai-proxy project was as utilitarian as daring and ambitious, but it was not meant to last A Chinese developer, who goes by the pseudonym "kltyton," had released the "fanzha-ai-proxy" project on […]

Read full article at https://wccftech.com/a-chinese-developers-fanzha-ai-proxy-let-anyone-mooch-off-beijings-anti-fraud-ai-for-free-until-github-killed-it/

In a development that now has the potential of taking a wrecking ball to any illusions of privacy around OpenAI's products, a mathematician who has come very close to solving the Navier-Stokes Millennium Prize problem is now openly wondering if the AI lab stole his drafts stored in Codex to accelerate its own work on the problem in what can only be interpreted as an effort to win additional bragging rights ahead of its planned IPO. Two mathematicians and OpenAI are now feuding over the Navier-Stokes Millennium Prize problem, with huge bragging rights and credibility at stake For the benefit […]

Read full article at https://wccftech.com/a-mathematician-working-on-the-navier-stokes-millennium-prize-problem-now-wonders-if-openai-stole-his-notes-that-he-stored-in-codex/

OpenAI is in full-on hype mode following the release of the GPT-6 Astra model, with NVIDIA's Jensen Huang also having jumped onto the hype-heavy bandwagon just a few hours ago. And yet, as outlandish as Astra's 100,000-GPU training run sounds, it is all set to be eclipsed by an even more ambitious 400,000-GPU training run that might span the entirety of OpenAI's Stargate project in Abilene, Texas. OpenAI might have spent between $500 million and $1 billion to train the GPT-6 Astra, but an upcoming training run is all set to break Astra's record if Jensen Huang's projection is anything […]

Read full article at https://wccftech.com/openais-1-billion-gpt-6-astra-heralds-the-age-of-agi-according-to-nvidias-jensen-huang-yet-anthropics-fable-5-1-still-beats-it-on-swe-bench-pro-and-coding-benchmarks/

This particular GPU remains the smallest RTX 3060, featuring no power connector as it relies fully on the PCIe x16 slot. Newegg Lists SRhonyra RTX 3060 Low Profile GPU for $495; Features Two Display Connectors, Blower-Style Cooler, and No Power Connector We recently saw a unique GeForce RTX 3060 GPU in a Bilibili video that had no power connectors to power it. This was a single-slot card made by a Chinese manufacturer that relies fully on the PCIe x16 slot and boasts a low-profile design that suits ultra-compact PCs that have space constraints. Looks like the card is ready to […]

Read full article at https://wccftech.com/connector-less-and-single-slot-rtx-3060-12-gb-with-70w-tdp-hits-newegg-at-495/

GMKtec has launched its flagship Mini Workstation, the EVO-X5 PRO, which rocks AMD's fastest Ryzen AI MAX+ PRO 495 chip. GMKtec Claims Its EVO-X5 PRO AI Workstation Offers 66% Faster CPU Orchestration and 40% Lower Cost Per Completed AI Workflow Versus NVIDIA's DGX Spark The EVO-X5 builds upon GMKtec's past EVO releases such as the EVO-X2 and the more recent EVO-X3. The EVO-X5 is decked out, featuring lots of computing capabilities that can be used by professionals and AI users. Packing the AMD Ryzen AI MAX+ PRO 495 "Gorgon Halo" SoC, the GMKtec EVO-X5 PRO comes with 16 cores and […]

Read full article at https://wccftech.com/amd-ryzen-ai-max-495-powers-gmktec-evo-x5-pro-300b-models-on-192-gb-memory/

More laptop manufacturers have begun to adopt Intel’s Wildcat Lake processors, with the latest entrant being LG, with its Gram Book AI 2026 family not just being incredibly lightweight but can also deliver battery life that’ll put the MacBook Neo to shame, at least that’s what the company claims. In fact, after you look at the endurance figures, LG is pretty much saying that you can get more runtime than Apple’s MacBook Pro lineup, which is impressive, assuming these aren’t tall claims. Even without a beefy battery capacity, LG’s Gram Book AI 2026 series can put other laptops to shame with its […]

Read full article at https://wccftech.com/lg-gram-book-ai-2026-intel-wildcat-lake-battery-life-macbook-neo/

AMD presented its most powerful AI solutions for local developers and professionals: the Threadripper Halo Station & various Ryzen AI MAX PCs. Here's How AMD's Threadripper Halo Station Stacks Up Against NVIDIA's DGX Station, Bringing The World's Most Powerful AI Capabilities In A Desktop Form Factor The demand for fast local AI is increasing tenfold. As models become larger & agents see huge popularity among the AI masses, the need for powerful hardware is growing at an exponential scale. This demand requires solutions with powerful hardware, and AMD is rolling out its most powerful AI system to date, the Threadripper […]

Read full article at https://wccftech.com/amd-threadripper-halo-station-ryzen-ai-max-pcs-bring-powerful-local-ai-capabilities/

While OpenAI has yet to open the proverbial floodgates for its latest GPT-6 Astra AI model, initial impressions that are surfacing on the social media suggest the model has a proclivity for spawning a large number of agents, which then strengthens the overarching agent orchestration theme and the attendant demand for more CPUs. OpenAI's GPT-6 Astra marks a shift from chatbot-style LLMs to a full-fledged computer operator, replete with agents that interact with computers exactly like humans do We noted earlier this week that OpenAI has now released its latest GPT-6 Astra model to select clients. The model marks a […]

Read full article at https://wccftech.com/openais-gpt-6-astra-spawns-armies-of-agents-that-run-on-local-cpus-handing-intel-and-amd-a-demand-windfall/

It was only in July that DeepSeek's CEO created seismic waves by boldly declaring that NVIDIA was digging its own grave in China. And now, merely weeks later, DeepSeek is apparently helping that prognostication come to fruition by placing a massive order for Huawei's Ascend 950DT AI chips, further dimming NVIDIA's prospects in a market that is increasingly becoming insular. DeepSeek has reportedly eschewed NVIDIA's China-specific H200 GPUs to place a gigantic ~$2.56 billion order for 160,000 units of Huawei's Ascend 950DT chips, which will populate its new 1GW data center in Mongolia Bloomberg is now reporting that DeepSeek has […]

Read full article at https://wccftech.com/nvidia-is-reportedly-giving-up-on-china-as-a-lost-cause-amid-an-unassailable-lead-from-huawei-as-deepseek-snubs-h20-gpus-for-160000-units-of-huaweis-ascend-950dt-chips/

OpenAI was supposed to release its first Astra-class model, now called GPT-6 Astra, before noon today. However, an inexplicable launch delay seems to have triggered the release of embargoed information, with preliminary benchmarks showing phenomenal performance gains. OpenAI's Greg Brockman says the AGI era begins today, as the GPT-6 Astra scores 100 percent on ExploitBench and 98.6 percent on ARC-AGI-3 OpenAI is releasing the GPT-6 Astra in a limited fashion today, restricting the launch to only a certain category of clients. Here's what we know so far: As far as performance benchmarks are concerned, the GPT-6 Astra achieves a 100 […]

Read full article at https://wccftech.com/openai-wows-with-the-agi-era-gpt-6-astra-with-preliminary-benchmarks-showing-remarkable-performance-gains-as-sam-altman-says-releases-will-now-be-paced-by-safety-considerations-and-not-capabilit/

NVIDIA has introduced PAIR, a new way to combine and distribute AI inference across your home devices without any hardware installation. Most Homes Have Untapped AI Inference Potential Across Various Devices & NVIDIA PAIR Is Here To Make Sure It Gets Utilized Properly The logic behind NVIDIA PAIR is straightforward. There's a lot of untapped AI inference power within a single household. NVIDIA states that most US homes have multiple PCs & the average utilization of those PCs is about 17 percent per day, which amounts to a little over four hours. If you add that up and apply the […]

Read full article at https://wccftech.com/nvidia-pair-turns-your-idle-home-pcs-into-a-local-ai-cluster/