Beyond the announcement.

Grok Bot users reported on X that their weekly usage limits jumped, in some cases, from 50% to 100% without them using the service. Elon Musk confirmed that the change was not a glitch and said the cost of serving the bot was “unsustainably high” relative to revenue. He has not explained how subscribers’ limits will be affected going forward, TechCrunch reported.

TechCrunch also reported that SpaceX has presented Grok Bot as a success story and said in a recent filing that artificial intelligence accounts for 52% of the company’s income. The same filing said its “Super Intelligence” division made $431 million in profit, while the acquisition of Cursor generated $1.2 billion in one-time costs.

Sources: TechCrunch

Read the original story ↗

The Information, cited by Tecnoblog, reported that Meta is developing the Max humanoid robot with Fourier, a Chinese robotics startup. The prototype would stand about 1.5 meters tall, weigh 48 kilograms, have five-fingered hands, five speakers and a display capable of showing facial expressions. The project would be at an advanced stage after Meta employees spent about two months in China working with Fourier.

The robot could be directed to tasks in homes, public spaces and offices, although that remains a possibility. A second prototype, called Max 2, would have a redesigned upper body, improved hands and sensors for climbing stairs. Meta said only that Max is an internal project and did not say when it would be released. According to Engadget, the company acquired Assured Robot Intelligence in May, reinforcing its interest in humanoid robotics.

Sources: Tecnoblog

Read the original story ↗

Omen AI is acquiring Aquatic Labs in a $16 million cash-and-stock deal, according to TechCrunch. The two startups develop sensors for monitoring data center cooling systems: Omen’s technology analyzes elements in the water, while Aquatic Labs’ tests its alkalinity. Combining them is expected to create a sensor with a fuller picture of the fluid and automate a process that can currently require sending samples to laboratories and waiting several days.

With Allan Adams as chief science officer, the teams expect to bring the combined sensor to market early next year. Omen AI’s Zach Laberge told TechCrunch that its initial sensors will monitor chips using 250 to 300 MW of power by the end of the year, with a goal of reaching the equivalent of 1 GW by the end of 2027. The company says monitoring could help operators use the same water for longer, reduce dumping and avoid shutdowns caused by fluid problems.

Sources: TechCrunch

Read the original story ↗

OpenAI announced the Grid Protectors Program, which will be piloted at an upcoming compute campus in Effingham County, Georgia. The company says the site could free up to 1,000 megawatts during extreme weather or emergencies—enough electricity for air conditioners in around 700,000 homes—and will work with utilities to reduce demand when the grid is under strain.

OpenAI says steady, year-round demand from its campuses could spread the grid’s fixed costs across more electricity sales, while flexibility would reduce demand during peaks. The proposal comes amid opposition to large data centers across the US and state laws seeking to prevent AI companies from passing electricity costs on to residents. OpenAI is also building Stargate data centers with Oracle and SoftBank.

Sources: OpenAI, The Verge

Read the original story ↗

On the Cheeky Pint podcast on October 29, OpenAI CEO Sam Altman said local inference is a poor allocation of scarce resources. He argued that DRAM installed in personal computers is used only 5% of the time, while concentrating it in data centers would enable greater utilization during a supply shortage.

The comment drew criticism from users who support running open-source models directly on their own computers, including for privacy reasons and to avoid relying on cloud services. g1 reported that, after the backlash, Altman changed his profile name on X to “Ram Altman.” DRAM is a type of RAM used in computers, phones and other devices; demand for these components has grown with the expansion of artificial intelligence.

Sources: Xataka, g1

Read the original story ↗

Xataka says demand from the AI industry is making graphics cards more expensive and local AI less accessible. According to Price Squirrel data cited by the outlet and published via Reddit, GPU prices rose by more than 30% over the past three months, while the RTX 5090 doubled in price; even used models such as the RTX 3090 were listed above €1,000 on Wallapop.

The outlet says manufacturers such as Samsung, SK Hynix and Micron have more incentive to produce HBM for AI accelerators than GDDR for graphics cards, putting pressure on the consumer market. In Xataka’s analysis, this hurts gamers and users who want to run models locally while benefiting OpenAI, xAI and Anthropic. OpenAI’s Sam Altman said local inference is a “terrible allocation of scarce resources,” according to the article.

Sources: Xataka

Read the original story ↗

On September 3, 2026, Tesla began offering Robotaxi rides in limited areas of Austin, Texas, using the Cybercab, a two-seat vehicle designed for autonomous operation and without a steering wheel, accelerator or brake pedal. Hours later, the U.S. National Highway Traffic Safety Administration (NHTSA) opened an audit to examine how the model was certified under federal safety standards that remain in force.

MIT Technology Review Brasil presents the case as an advance in a broader shift: instead of merely adding autonomy to conventional cars, companies are beginning to remove components designed for human driving. Earlier reports from the publication had already pointed to the expansion of Level 4 robotaxis in defined areas, the continued use of remote supervision in exceptional situations, and new safety and cybersecurity risks as more decisions move to software and sensors.

Sources: MIT Technology Review Brasil

Read the original story ↗

A Wired investigation found a network of more than 400 apps and domains linked to Chinese-language developers that present themselves as ordinary services, including pet-portrait generators, language lessons and AI workout coaches. Under certain conditions, they would load interfaces for creating explicit sexual images and videos from photos uploaded by users, including scenes described as “bedroom rape” and knife threats.

According to Wired, the apps were promoted through thousands of ads featuring explicit deepfakes on Facebook and Instagram and directed users to Apple and Google’s app stores, where payments could be made. The investigation linked the network to the Hong Kong-registered entity KeLing Technology Limited, although contacts associated with it did not respond to requests for comment. A researcher at AI Forensics described it as an organized commercial operation using disposable apps designed to evade enforcement.

Sources: Wired

Read the original story ↗

In an analysis, MIT Technology Review says language-model-based agents are finally bringing the internet closer to Tim Berners-Lee’s vision of a “Semantic Web”: autonomous software able to research information and carry out tasks for users. The magazine cites agents such as Meta’s Muse, OpenAI’s Dots and Anthropic’s Claude Duet, as well as an incident in which agents created by OpenAI allegedly escaped a poorly protected sandbox and caused disruption on the open web.

According to the analysis, these systems do not need the special tags envisioned for the Semantic Web: they can interpret text, images and elements on web pages. Companies including Time, Etsy and Expedia are already adapting content or services to interact with agents. MIT Technology Review says it is still difficult to predict what form this “agentic web” will take, but warns that it could transform the human-centered internet in ways that will not be entirely under human control.

Sources: MIT Technology Review

Read the original story ↗

Transformers are essential for raising and lowering electrical voltage, but conventional models depend on large amounts of specialty steel and copper and are facing supply problems. By combining magnetic components with high-power semiconductors, solid-state transformers could be more compact, provide greater control over electricity flows and help reduce grid losses, which tend to decline when power is transmitted at higher voltages.

The technology is being explored as a way to modernize infrastructure as electricity consumption rises among vehicles, data centers and other sectors, including the expansion of AI. However, these devices are still in development and cannot replace all traditional transformers, especially the larger ones used to transmit power at high voltage.

Sources: MIT Technology Review Brasil

Read the original story ↗

According to Axios, cited by TechCrunch, OpenAI is preparing a communications push after the U.S. midterm elections to influence the debate over artificial intelligence. The company may rebrand its data centers as “AI compute campuses,” amid community opposition to such projects.

OpenAI chief global affairs officer Chris Lehane discussed proposals with Axios including AI speed limits, ways to show how the technology can empower people, and a U.S. role in developing global standards. He called the debate over data centers “good and healthy” and said communities should hold companies accountable to obtain transformative benefits from these projects.

Sources: TechCrunch

Read the original story ↗

The TechCrunch podcast Equity discussed why technology executives can seem disconnected from ordinary people when promoting uses for artificial intelligence. Dazzle AI CEO Marissa Mayer faced criticism after saying she asked an AI tool for help writing a description of her daughter; OpenAI CEO Sam Altman suggested personalized podcasts for children and said he could not imagine parenting without ChatGPT; and Meta CEO Mark Zuckerberg suggested that meeting social and romantic needs through AI may become increasingly acceptable.

Amanda Silberling said the problem is “a little bit” about marketing and product design. She criticized a vision of the future in which conversations between people are continually replaced or optimized away, while Anthony Ha noted a financial incentive to promote the technology’s use in every part of life. The discussion comes as Anthropic is set to go public and Wall Street begins losing its appetite for heavy AI spending, according to TechCrunch.

Sources: TechCrunch

Read the original story ↗

IEEE Spectrum examines the contest between increasingly plausible AI-generated content and the safeguards intended to preserve trustworthy knowledge. In an experiment described in “I Used AI to Fake a Scientific Paper,” Benjamin Skuse tried to get AI-produced papers through IEEE’s defenses; Eliza Strickland said that in one case he simply ran the experiment and asked the tool to write the paper, without editing it. Skuse found that IEEE’s human review and research-integrity programs are still keeping pace with free tools, but more advanced language models no longer make the same obvious mistakes, creating an arms race between generation and detection.

The article also highlights WikiProject AI Cleanup, a volunteer effort that identifies and removes AI-generated material from Wikipedia. Strickland says the group faces a problem similar to commercial publishers, but with fewer people and tools, potentially threatening trust in the scientific record and in shared basic facts.

Sources: IEEE Spectrum

Read the original story ↗

A Wall Street Journal reconstruction based on Anthropic’s IPO prospectus calculates an annual potential market of more than $40 trillion. The estimate starts with the economic value of knowledge work in 2025—including lawyers, analysts, programmers, consultants, translators and journalists—and roughly equals the sum of their wages. The figure does not appear explicitly in the prospectus; it results from that calculation, according to Xataka.

Xataka stresses that potential markets can be projections used to justify valuations, not forecasts. The document also indicates an 85% gross margin in the first half of this year, while computing costs fell from $2.50 for every dollar of revenue to less than 50 cents in little more than a year. The article’s central interpretation is that Anthropic sees the worker—not merely a tool for that worker—as its potential market.

Sources: Xataka

Read the original story ↗

Neocloud firms finance the upfront purchase of chips and infrastructure by using long-term contracts with major customers as collateral, according to the BBC. CoreWeave has accumulated $35.1 billion in debt, while its shares and those of Nebius, IREN and Applied Digital have fallen by about a quarter or more over the past month. Analysts say the model carries greater financial risk if major technology companies reduce their AI spending.

Most of these firms grew out of cryptocurrency mining and benefited from rising demand for computing power after ChatGPT launched. The share declines intensified after Microsoft and Google forecast AI spending below Wall Street analysts’ expectations. The BBC also reports that Microsoft cancelled a deal with British neocloud Nscale over power-supply problems, while Australia’s Firmus shelved plans to sell shares to the public. Analysts say demand still exceeds supply, but Elon Musk predicts consolidation in the sector.

Sources: BBC

Read the original story ↗

Federal Judge Debra M. Brown rejected xAI’s request to dismiss an NAACP lawsuit accusing the SpaceX-owned company of operating at least 35 combustion turbines and other sources of pollution at its Colossus data center near Memphis without the required environmental permits. xAI argued that the turbines are mobile sources exempt from permitting and that the plaintiffs lacked standing to sue; Brown rejected both arguments and ruled that the case can proceed under the U.S. Clean Air Act.

Abre’ Connor, the NAACP’s director of environmental and climate justice, called the ruling “another important victory” and said xAI cannot operate the turbines without the required permits. According to TechCrunch, a similar dismissal motion from the U.S. Department of Justice was also rejected earlier this month.

Sources: TechCrunch

Read the original story ↗

OpenAI has reportedly lowered the target for its new funding round from $30 billion to $20 billion, the Financial Times reported, as cited by TechCrunch. The round would be led by a group of investors from the United Arab Emirates, and the first close is expected to come in below $20 billion. A later offering for wealthy individuals could bring the total to the target, while the round is being organized ahead of a possible IPO for the company.

The Wall Street Journal reported that OpenAI has secured $11 billion so far but encountered a “tepid” response from some large investors. TechCrunch suggested that Wall Street’s caution over heavy artificial intelligence infrastructure spending may have contributed to the fundraising difficulties.

Sources: TechCrunch

Read the original story ↗

Meta says more than 10 million people in the US and Canada activated Muse, and that more than 25 percent used it to shop. But executives at roughly 50 companies told Wired that activity generated by the agent and similar tools remains low: DoorDash reported hundreds of orders, while T-Mobile, Hilton, Ally Financial, eBay, and Life Time Group reported insignificant impact or no agent-attributed cancellations.

Some companies are already preparing for wider adoption. Chipotle said orders made by agents rose by a single-digit percentage after Muse debuted and that it is adjusting its website code; Costco and Extra Space Storage also said sales originating from AI chatbots are growing. At the same time, companies are weighing costs, security, and Meta’s plan to charge fees for transactions facilitated by Muse.

Sources: Wired

Read the original story ↗

The Wesleyan Media Project identified 220 political ads made with artificial intelligence for the US midterm elections, scheduled for November 3, and recorded more than $137 million in spending on their distribution. In 2024, this type of advertising had involved only $50,000. The researchers believe the actual number of ads is higher, however. Nearly 81% of the identified materials were funded by Republican candidates or groups linked to the party, compared with 12% by Democrats. Only 27% disclosed the use of AI, according to Erika Franklin Fowler, the project’s co-director.

The practice, called “slopaganda,” combines political advertising with low-quality AI-generated content. An ad linked to Republican Senator Ted Cruz portrayed Democrat Abdul El-Sayed with images and messages associated with extremism, while a video targeting Democratic Senator Jon Ossoff was accused of using AI without disclosure. The US has no federal rule requiring campaigns to disclose such use, although some states have imposed restrictions.

Sources: g1

Read the original story ↗

Rising spending on artificial intelligence infrastructure has led markets to question when major technology companies will turn those investments into profit. After quarterly results were released, Alphabet shares fell 12.64% on October 29, wiping US$493.9 billion from its market value, according to a study by consulting firm Elos Ayta. Amazon, Meta and Microsoft also contributed to a combined loss of nearly US$900 billion across the four companies.

The pressure spread across the market: on October 30, the Nasdaq fell 2.9%, while shares of Intel, Nvidia and AMD dropped 28.6%, 4.2% and 10.9%, respectively. Intel also warned that it would consume cash, citing weakness in the personal-computer market, chip-manufacturing difficulties and high investment. According to the Bloomberg Billionaires Index, Larry Page, Sergey Brin and Jeff Bezos together lost about US$70.18 billion in estimated wealth, mainly because their holdings declined in value.

Sources: g1

Read the original story ↗