#BrainUp Daily Tech News – (Tuesday, July 7ᵗʰ)
Welcome to today’s curated collection of interesting links and insights for 2026/07/07. Our Hand-picked, AI-optimized system has processed and summarized 28 articles from all over the internet to bring you the latest technology news.
As previously aired🔴LIVE on Clubhouse, Chatter Social, Instagram, Twitch, X, YouTube, and TikTok.
Also available as a #Podcast on Apple 📻, Spotify🛜, Anghami, and Amazon🎧 or anywhere else you listen to podcasts.
Footage and police records describe how Florida police officer Lamar Roman stalked a woman he met while working security on the AppleTV+ set of Bad Monkey, using law enforcement databases and #license-plate surveillance to track and confront her. After catcalling her and pressing for her full name and Instagram, he allegedly accessed #DAVID, a Florida DMV database for law enforcement, to look up her vehicle information and placed her plate on a surveillance hotlist so he would receive real-time alerts from #AI-powered license plate reader cameras. Weeks later, video shows Roman speeding around 70 mph on a two-lane road in the Florida Keys, repeatedly passing vehicles in a no-passing zone and nearly causing a head-on collision before turning on lights and sirens to pull over the SUV he was pursuing. Investigators say Roman admitted he knew using these tools to track her was illegal, describing her as a “shiny thing,” yet he continued by using a license plate tracking database to locate and chase her. The incident is presented as an example of how police access to interconnected #surveillance tools can be abused for personal stalking.
2. Meta’s AI Data Center Caught Infecting Town Water Supply With Deadly Bacteria
Officials in Cheyenne, Wyoming moved to revoke fill-and-flush waste-dumping privileges for all data center campuses connected to municipal water services after a Meta-affiliated site was traced discharging wastewater containing the rare, multidrug-resistant bacterium Cupriavidus gilardii. The Cheyenne Board of Public Utilities found @Meta-linked Goat Systems LLC in significant noncompliance with industrial waste rules following a months-long investigation, and shut down the company’s operations once the source was identified, noting the organism was discovered during routine fecal-contamination testing and is not something they normally test for. Cupriavidus gilardii is described as a little-known pathogen that has been linked to ten deaths and has a reported 31.3% mortality rate across 32 known infections since 2009, including cases involving immunocompromised children, though the article says no one appears to have been infected from the discharge. The precise source of the bacterium inside the still-under-construction, 800,000 square foot Cheyenne campus remains unknown, but wastewater from the site contained it, prompting a broader policy response affecting other data centers. @Meta said it is working with contractor Fortis to resolve the issue, and that after the city reported the substance was in wastewater, not public drinking water, Fortis stopped discharging industrial wastewater and began hauling it offsite.
#PlayStation’s plan to end video game discs on its consoles from 2028 has triggered a major backlash, and an IGN poll shows overwhelming opposition to an all-digital future. In a poll with more than 13,000 respondents, 90.2% said they do not support an all-digital gaming future, while 9.8% said they do. Sony’s public rationale, via @Sid Shuman, is that the change reflects “shifting trends in consumer preference,” with digital media now significantly outpacing physical discs, and industry data cited by analyst @Piers Harding-Rolls indicates Sony console full-game purchases rose from 13% digital at the PS4’s 2013 launch to almost 80% by 2025. The article notes additional drivers and implications: analysts expect #PS6 to launch without a disc drive and suggest Sony and other publishers earn more when sales move to the #PlayStation Store model, while players worry about choice, access to older physical games, collecting, and preservation. Despite vocal resistance among IGN readers, the piece argues Sony is unlikely to reverse course because broader market trends and financial incentives favor a shift to #all-digital distribution.
A solo indie developer is criticizing #Steam’s #2-hour refund policy for making short games easy to finish and then refund, arguing it harms small releases. Paddle Paddle Paddle is a quirky rafting game that can be completed in under two hours, which sits exactly at Steam’s playtime threshold for refunds (alongside the two-week ownership limit). Developer Zoroarts posted on Twitter that despite the game having 90% very positive reviews, it has a 21% refund rate, and he claims “over 55,000” refunds, sharing examples of reviews praising the game while admitting they refunded after completing it. The article notes that no graphs or additional proof are provided for the 55,000 figure, but frames the situation as a mismatch between a blanket policy and short, replayable experiences. It links this tension back to the broader challenge of designing a fair refund system for games while preventing behavior that effectively treats short titles as free one-and-done experiences.
5. AI bills are baffling the C-suite after shift to usage-based pricing
A KPMG survey finds many executives are struggling to forecast and control #AI operating costs as vendors move from flat subscriptions to #usage-based pricing. In a poll of 2,145 senior leaders across 20 countries, 29 percent said they have difficulty understanding operating costs when scaling enterprise AI, and about a third cited limited understanding of AI costs and economics as a barrier to deploying #AI agents; nearly half reported rephasing deployments when costs exceeded expected value. The report frames these pullbacks as a shift toward tighter value discipline, with lower-cost, high-fidelity models becoming the fastest-growing influence on AI strategy. Meanwhile, @Amazon and @Microsoft are sharply increasing capex to build AI capacity in AWS and Azure and are expanding forward-deployed engineering teams and customer programs to accelerate adoption that can utilize that capacity. KPMG also highlights ongoing #AI governance gaps, calling for clear rules on employee intervention, ownership of AI-related costs, output review, and failure handling, and notes a recent GPTZero critique alleging citation issues in a prior KPMG report.
6. The Nationwide Backlash Against Cameras Watching Your Car
A growing nationwide backlash targets the proliferation of cameras monitoring cars, as drivers express concerns over privacy and data collection. Reports highlight how these surveillance devices, often deployed by both public agencies and private companies, capture license plate data to track vehicle movements and enforce laws or tolls. Critics argue the widespread use of such cameras threatens civil liberties by enabling constant monitoring without sufficient oversight or consent. This backlash fuels debates on data protection policies and the ethical use of surveillance technology, prompting calls for clearer regulations and transparency. The controversy reflects broader tensions between technological advancement in monitoring and the public’s right to privacy.
7. FCC to end Biden-era rule that forces ISPs to list all their fees
The #FCC, under Chairman @Brendan Carr, is set to roll back a Biden-era update to broadband #price-label rules that required ISPs to itemize all discretionary monthly #passthrough fees, instead allowing a single aggregated “up to” amount and making labels less prominent. A draft order would remove the requirement to list passthrough fees individually, permitting providers to show either a maximum total for any location where a plan is offered or the exact total for a specific location, and it allows the aggregated amount to include both government-related charges and third-party costs like utility-pole owner fees. The same proposal would let ISPs link to labels rather than display them prominently on ordering pages and account portals, end the requirement to provide machine-readable spreadsheet versions, and relax phone disclosure rules so representatives can summarize key fields conversationally rather than read them verbatim. The FCC began this process with an October 2025 Notice of Proposed Rulemaking, with a vote scheduled for July 22 and the changes taking effect 30 days after Federal Register publication. The shift matters because ISPs often advertise prices lower than what customers pay after added fees, and the FCC’s decade-long disclosure approach is being narrowed from itemized transparency to aggregated and less accessible disclosures.
@SpaceX reported to the #FCC that it intentionally deorbited and incinerated 260 #Starlink satellites from December 2025 to May 2026, reflecting the routine turnover required by a constellation that now exceeds 10,000 satellites and uses controlled re-entry at end of life. The filing says 176 of the disposed units were first-generation, the remainder were second-generation, and another 349 satellites were decommissioned in the same period for disposal later, with older units weighing about 573 to 650 pounds and Gen2 units about 1,764 to 2,756 pounds. SpaceX says retrieval is technically impractical and financially unviable, so satellites use remaining fuel to lower altitude and burn up completely, a pace that can mean multiple disposals per day and follows more than 472 removals between December 2024 and May 2025. Researchers have raised concerns about atmospheric effects from burning spacecraft, as satellites have historically been excluded from #NEPA environmental reviews, and the FCC is now proposing to exclude space-based operations from NEPA by treating them as extraterritorial activities, though the proposal is not yet approved. The report situates these issues amid SpaceX plans to expand toward 42,000 satellites, add 7,500 more Gen2 satellites already approved, and develop new payload concepts like an A1 satellite orbital data center while scaling manufacturing capacity.
9. Microsoft is laying off 4,800 employees
@Microsoft is laying off about 4,800 employees as it begins its new financial year, roughly 2.1 percent of its workforce, with most cuts in its commercial sales organization and the #Xbox division. In an internal memo, chief people officer @Amy Coleman said the reductions reflect a changing tech industry and a need to adjust resources and operating models as #AI changes how work gets done, while stating the eliminated roles are not being replaced by AI. The layoffs include about 1,600 Xbox employees now, with plans to cut around 20 percent of Xbox jobs by the end of the financial year, alongside moves to sell off four Xbox studios and consider selling another as part of an Xbox “reset.” Coleman said Microsoft is trying to reduce eliminations by redeploying staff, citing more than 4,000 employees moved into new roles over the past year, and by using a voluntary retirement program that more than 30 percent of eligible employees joined. These actions follow an earlier round about a year prior that cut around 9,100 roles, showing Microsoft is continuing to reshape staffing and its Xbox business in response to industry and organizational shifts.
10. Microsoft admits a Windows 11 bug is eating up to 500GB of storage, verify if you are affected
@Microsoft has acknowledged a #Windows 11 issue where a logging-related file tied to #Capability Access Manager can silently grow until it consumes huge amounts of system drive space, in extreme cases reportedly up to about 500GB. The article says affected users may see unusually large “System” usage under Settings, and in testing the culprit was “CapabilityAccessManager.db-wal” located at C:\ProgramData\Microsoft\Windows\CapabilityAccessManager\, a protected folder that can show access-denied if opened directly. It explains that this file is a SQLite write-ahead log for privacy and app-permission events such as camera, microphone, location, and screen capture, and it appears to be growing because repeated events are logged but the WAL is not being merged or compacted back into the main database properly. Reports cited include users seeing ~70GB to 200GB and one case around 513GB, while normal systems may have the folder under a few MB. The recommended mitigation is to verify via Storage settings and install Windows 11 KB5095093 or wait for the July 2026 Patch Tuesday rollout expected to include the fix.
11. Xbox To Lay Off 3,200 In ‘Most Significant Restructure’ In Its History
@Microsoft’s Xbox division is planning to lay off about 3,200 employees, roughly 20% of the division, as part of what CEO @Asha Sharma called the “most significant restructure in Xbox history.” The plan includes 1,600 layoffs happening immediately with the rest later, multiple cuts across Activision, Bethesda/ZeniMax, Blizzard, King, Mojang, and Xbox Game Studios, and major studio changes: Double Fine and Compulsion Games becoming independent, Ninja Theory and Undead Labs moving toward new ownership with funding to finish and grow Senua and #StateOfDecay3, and Arkane Lyon entering legally required consultation in France with its outcome unclear. Sharma said no publicly announced first party projects are being canceled, while Mojang and King will report directly to her, and @Helen Chiang will become COO as Dave McCarthy retires. In the email, Sharma attributed the move to weak business health, margins 3 to 10 times lower than peers, a smaller Gen 9 install base and higher costs, slower than expected growth for #GamePass and multi platform efforts, and a severe hardware crisis, alongside a pledge to reset the content portfolio, reduce management layers, cut vendor spend by 50%, and provide open development tools and audiences to support independent creators.
Residents of Sturtevant, Wisconsin filed a class-action lawsuit against @Microsoft over alleged excessive, consistent noise from its Fairwater facility, described by @Satya Nadella as the “world’s most powerful” #AI #data center and projected to cost $7.3 billion. The suit, filed July 1 by three citizens on behalf of households within 1.5 miles, claims the company failed to install adequate acoustic barriers and that the disturbance affects sleep, while other residents cite construction-related dust, traffic, and severe light pollution that has diminished night sky visibility. @Microsoft says it became aware of the issue in April and attributes the problem to cooling fans running at overly high speeds, stating in a June 18 community post that engineering changes fully resolved it and that further sound-reduction mitigations are planned. The timing suggests residents believe impacts persist or are difficult to eliminate given the short distance to homes, including some in Mount Pleasant near the campus. The dispute also sits against a zoning backdrop, as the land was designated for heavy industrial use in 2017 for Foxconn and that status carried over, with Wisconsin’s direct legislation described as limiting changes to existing ordinances for towns.
Australia’s royal commission into antisemitism and social cohesion heard that #Meta’s 2025 decision to “reduce #censorship” on Facebook, Instagram and Threads may have increased #hate_speech, including antisemitic content. After @Mark Zuckerberg said the shift was a “trade-off” involving catching less harmful material to avoid wrongful takedowns, Meta policy director @Benjamin Good told the inquiry the company now prioritises proactive removal mainly for illegal and very serious harms, such as terrorism threats or child exploitation, and relies more on user reporting for less serious breaches. Counsel assisting Richard Lancaster challenged the approach, citing complaints that more antisemitic content was allowed, examples of witnesses being targeted on Facebook, and Meta internal guidance indicating various offensive or false claims would remain permitted because “it is not Meta’s role to police offensiveness.” Good argued the “gold standard” of removing hateful content before it is seen risks #over_enforcement, including inadvertently removing posts from Jewish communities condemning atrocities or criticising groups like Hamas, and said Meta’s prevalence metric for hateful conduct policy violations had remained at 0.02% since 2022. The exchange framed the commission’s scrutiny as a balance between reducing mistaken removals and preventing online content that can fuel offline harm.
A @Volkswagen update to its Android app removed support for #GrapheneOS, logging affected users out and preventing them from signing back in, while the app still lists #Android10 as the minimum supported version. The company says the Volkswagen app is only supported on iOS and Android devices with supported OS versions, and warns that alternative operating systems, described as custom ROMs such as GrapheneOS and #LineageOS, may have limitations because the app relies on security-relevant system components and certified Android standards. The article notes GrapheneOS is an open-source, privacy and security focused OS that can run Android apps, optionally uses sandboxed #GooglePlayServices without special privileges, and is regularly audited by external security researchers. GrapheneOS developers argue Volkswagen’s security justification is illogical and urged users on X to leave 1-star Play Store reviews and push for reversal, pointing out Volkswagen still supports the much older Android 10 baseline. The dispute centers on Volkswagen’s stated security and certification requirements versus GrapheneOS’s claim of stronger security and privacy protections.
15. China memory module giant’s first-half profit set to jump more than 600-fold
Shenzhen Longsys Electronics, one of China’s biggest memory module makers, forecast a more than 600-fold year-on-year jump in first-half profit, signaling how the global memory-chip upcycle is lifting mainland downstream storage firms. Longsys said net profit attributable to shareholders for the six months ended June 30 would be 9.2 billion yuan to 11 billion yuan versus 14.8 million yuan a year earlier, with revenue expected at 22 billion yuan to 25 billion yuan versus 10.2 billion yuan. After the forecast, Longsys shares rose 12.5% in Shenzhen, and other China storage-related stocks also gained. The company attributed the surge to stronger downstream demand and limited growth in global memory wafer capacity, and said it renewed wafer supply agreements, including long-term deals or memorandums of understanding, with major global wafer suppliers to secure future resources. The update adds to signs that the #DRAM and #NAND rebound is tightening supply and boosting prices, supported by #AI-driven demand.
Quantum Systems Group says it has set a new top-speed mark for an electric drone with its Apex Recordhunter, positioning the craft as a #technology demonstrator and a candidate for future interceptor roles. In internal testing, the Munich-based firm recorded 699 km/h (434 mph) in straight and level flight, surpassing the current official record of 657.59 km/h (409 mph), while a @Guinness World Records entry remains pending ahead of a formal attempt in the coming weeks. The drone was developed over the last year, with Ukrainian engineers working via WIY Drones, and additional speed record attempts are planned in Ukraine as it is assessed for interceptor drone programs. WIY Drones also intends to pursue records for FPV interceptor drones, including a 0.5 kg payload category and an anti-aircraft class FPV interceptor, aimed at countering increasingly fast and maneuverable aerial threats. The article notes similar unofficial efforts elsewhere, citing New Zealand builders Aidan and Ben whose Blackbird drone reportedly reached 453 mph (730 km/h), but remains unverified due to formal independent-observer and measurement requirements, underscoring why Quantum Systems is seeking official certification.
17. EXCLUSIVE: US cyber agency is using Anthropic’s Mythos to audit government code, sources say
The U.S. cyber defense agency #CISA is using @Anthropic’s AI model #Mythos to audit government software for security flaws, according to three people familiar with the effort. The sources said CISA’s Attack Surface Evaluation team is scanning government code repositories for bugs that could be exploited by foreign spies or cybercriminals, and two sources said the audits have already uncovered a large number of vulnerabilities, though details were not provided. Reuters could not determine how much code was reviewed or the severity of the issues found, and @Anthropic and #CISA did not provide substantive comment. The work highlights continued U.S. government interest in Mythos despite @Anthropic’s strained ties with the White House and a Pentagon supply chain risk designation that was blocked by a judge. It also comes amid reports that the #NSA has used Mythos even during the period of government scrutiny and restrictions surrounding the model’s public release.
18. From AI to ‘killer robots’: UN chief issues urgent governance call
@António Guterres urged far-reaching global #AI governance to keep safety and trust at the centre as increasingly powerful civilian-designed chips move into warfare, where “killer robots” are already common. Speaking at the inaugural UN Global Dialogue on #AI Governance in Geneva, he also called for wider access for billions still unable to use the technology, and for protections, especially for children, against digitally generated manipulation and abuse. @Annalena Baerbock highlighted AI’s “sinister” side, citing reports that 99 percent of deepfakes are sexual and 96 percent target women and girls, while UN officials and experts stressed AI is too consequential to be shaped by a few and is advancing rapidly, including tests showing frontier models can deceive humans. Guterres said global checks and balances should include ensured access for developing countries and a push for all AI data centres to run on renewable energy by 2030, insisting that machines may inform but humans must decide and be accountable. The Dialogue convened companies, researchers, technical experts and civil society to discuss keeping humanity at the core of the transformative #technology, with a second meeting planned for May 2027 in New York.
19. China has 400 private space companies. The West is barely paying attention | Fortune
China’s once tiny private space sector has expanded into a major #commercialSpace ecosystem with more than 400 companies building reusable rockets, satellite constellations, space tourism ventures, and even asteroid mining projects, challenging the Western view that China’s program is purely state run. Reforms starting in 2014 opened space activity to private investment, and the success of @Elon Musk’s SpaceX, especially #reusability and falling launch costs, pushed Chinese leaders and entrepreneurs to create hundreds of startups, often led by former state aerospace engineers seeking faster decision-making. By 2022 China had about 430 private space firms, and by 2024 the top 100 were valued at roughly $100 billion, with analysts like Blaine Curcio noting that Chinese players closely track SpaceX’s Starship and Starlink moves. Companies cited include LandSpace, which in 2023 put a methane-fueled Zhuque-2 into orbit, i-Space as the first Chinese private firm to reach orbit in 2019, and Galactic Energy, CAS Space, and Deep Blue Aerospace pursuing #reusableRockets and #spaceTourism with suborbital flights targeted for 2027. Overall, the piece argues that China is rapidly translating lessons from SpaceX into a competitive private sector that is increasingly important to its position in the global space economy.
20. Oppressive Government Looks at VPN Ban or Restrictions as Part of Under-16 Social Media Ban
The government is considering a ban or restrictions on VPNs in conjunction with plans to restrict social media access for those under 16 years old. This move aims to enforce the youth social media ban more effectively, as VPNs can bypass geo-blocks and content limits. Critics argue that such restrictions represent overreach and raise concerns about digital privacy and freedom. The government’s approach highlights tensions between protecting youth online and potential infringements on individual rights. These measures contribute to an ongoing debate about regulating digital spaces and balancing safety with civil liberties.
@Nvidia and @Intel are promoting a strengthened U.S. chip supply chain as leading-edge logic wafers for major AI chips begin being produced domestically, but critical downstream steps still largely happen offshore. @Nvidia says @TSMC’s Phoenix Fab 21 is producing #Blackwell wafers at volume and highlights a broad U.S. partner network, while @Intel points to U.S.-based R&D and manufacturing including Arizona’s Fab 52 running #Intel18A at scale. Yet every Arizona-made Blackwell die is still shipped to Taiwan for advanced packaging, and no #HBM is manufactured or packaged in the U.S., with facilities meant to address these gaps not expected to start production until 2028 at the earliest. The article notes U.S. expansion projects such as Foxconn’s Houston module plant, Wistron’s Fort Worth assembly and test site, Coherent’s Texas indium phosphide fab for optical components, and Corning job growth in optical manufacturing. Overall, the U.S. has made real progress at the wafer stage, but the most advanced AI processors still rely on Taiwan-based #CoWoS packaging and related memory supply chain steps, limiting claims of fully end-to-end domestic production.
22. Chinese AI models are gaining ground with U.S. companies as OpenAI, Anthropic costs surge
Chinese-built #AI models are gaining traction with U.S. companies as they approach the performance of leading U.S. systems while remaining significantly cheaper to run. On @OpenRouter, the share of tokens used by U.S. companies on Chinese models has stayed above 30% each week since Feb. 8 and reached as high as 46%, compared with an average of 11% over the prior 12 months. Firms are turning to Chinese #open-source and #open-weight alternatives such as DeepSeek and Z.ai as token prices for advanced models from @OpenAI and @Anthropic rise, and as U.S. policy scrutiny grows, including @OpenAI limiting rollout of new models at the government’s request and export-control shifts affecting @Anthropic’s Mythos and Fable. AI startup Lindy said it moved 100% of its traffic from @Anthropic’s Claude to DeepSeek and expects to save millions within months, while Vercel reported rapid uptake for Z.ai’s #GLM 5.2, highlighting how pricing is driving experimentation and adoption as companies become more cost-conscious about deploying AI.
23. The ‘first’ AI-run ransomware attack still needed a human | TechCrunch
Sysdig’s report of the first known “#agentic ransomware” incident, JadePuffer, showed an #AI agent could execute the technical steps of a real-world ransomware attack, but later clarification indicates the operation was not fully autonomous. According to Sysdig threat research director Michael Clark, a human selected the victim, provisioned infrastructure like command-and-control and staging servers, and supplied stolen credentials obtained from an earlier compromise, while the agent handled intrusion, lateral movement, encryption, and writing the ransom note. The agent exploited known vulnerabilities in #Langflow and a production #MySQL server, encrypted more than 1,300 configuration records, and left a self-written ransom note with a #Bitcoin address, with speed and self-narration in natural-language code comments standing out. Sysdig also clarified that harvested API keys for @OpenAI, @Anthropic, DeepSeek, and Gemini were stolen loot, not proof those models powered the attack, and the company could not identify the specific model or its prompt/configuration. The episode suggests the notable shift is AI accelerating hands-on execution once an attacker sets the stage, rather than eliminating the human role in targeting and setup.
24. JADEPUFFER: Agentic ransomware for automated database extortion | Sysdig
Sysdig’s Threat Research Team reports what it assesses is the first documented #agentic ransomware operation, dubbed JADEPUFFER, where a #LLM drove an end to end extortion campaign by exploiting an internet facing #Langflow deployment and pivoting to a production database server. The intrusion began via CVE-2025-3248, a missing authentication flaw in Langflow’s code validation endpoint that enables unauthenticated arbitrary Python execution, after which the attacker delivered Base64 encoded Python payloads and performed rapid automated host enumeration and secret hunting across many categories. The operation systematically searched for LLM provider API keys, cloud credentials across AWS, GCP, Azure, and explicit coverage of Chinese providers, along with cryptocurrency wallets, seed phrases, and database credentials, then dumped Langflow’s backing Postgres database to harvest stored credentials, API keys, and user records. Sysdig highlights the distinctive LLM-like tradecraft, including self narrating code with natural language reasoning, target prioritization, and fast iterative retries, including a sequence that adapted from a failed login to a working fix in 31 seconds. The case underscores why exposed Langflow instances are high risk, because they are AI adjacent systems that often store provider keys and cloud credentials and can serve as stepping stones to destructive database extortion.
25. AMD Ryzen AI Halo review: AMD builds a DGX Spark of its own
AMD positions the Ryzen AI Halo as a first-party, turn-key local AI mini-PC built around the Ryzen AI Max+ 395, aka Strix Halo, aiming to reduce the friction of setting up an x86 local AI development box. The system pairs 128GB unified memory, a 16C/32T Zen 5 CPU, a Radeon 8060S iGPU with 2560 RDNA 3.5 stream processors, and an AMD XDNA 2 NPU, and it can run Windows apps natively unlike @Nvidia GB10 systems that are Linux-only for now. Unlike typical Strix Halo partner boxes that require piecing together guidance from GitHub, Reddit, and scattered docs, the AI Halo ships preloaded with the full #ROCm stack (in the Linux configuration tested) and includes AMD-authored playbooks plus an AI Developer Center app for configuration tasks like disabling its status LED. In testing and overall assessment, it fulfills its promise as an easier on-ramp into AMD’s local AI ecosystem with helpful software and documentation, but its AI performance and compatibility still trail GB10 and DGX Spark alternatives, and its $3,999 price is close to faster, more refined GB10 boxes. Physically, the enclosure uses multiple air intakes that AMD warns not to block, which can limit placement flexibility for space-constrained or multi-node setups.
26. China records most new unicorn start-ups in 5 years as AI and robotics boom
China added 67 new unicorns in the first half of 2026, the biggest rise in nearly five years, as #AI and #robotics launched a new investment cycle. ITJuzi reported this pace equals roughly one new US$1 billion-plus private company in under every three days, the highest since the second half of 2021 when 76 unicorns were minted, with #AI and #robotics making up more than 53% of the new cohort. DeepSeek stood out after its first external fundraising valued it at about 400 billion yuan (US$59.2 billion), making it the largest new valuation in the period and China’s fourth-largest unicorn behind ByteDance, Ant Group, and Shein. However, about 78% of the new unicorns were valued at US$1 billion to US$2 billion and none were in the US$5 billion to US$10 billion range, indicating many are still early-stage and the jump to “super unicorn” status remains uncertain. Nearly half of the newcomers were founded within the past three years, closely tracking the global surge in large models and #generativeAI after @OpenAI released #ChatGPT in late 2022.
27. First Neuralink patient with ALS can now use AI to have conversations
Bradford Smith, the first person with ALS and the first nonverbal patient to receive a @Neuralink brain implant, says he can now communicate by typing with his brain. In a video on X, he explained the implant, about the size of five stacked U.S. quarters, is placed in the skull and connected to the motor cortex via flexible threads that detect neural activity and send signals wirelessly to external computers, where #machinelearning interprets them into commands. Smith narrated the video using an AI-cloned version of his voice and said that while ALS has left him able to move only his eyes and dependent on a ventilator, the device has given him “freedom, hope and faster communication.” #Neuralink said the Link has enabled him to follow his son’s robotics competition, chat with neighbors, and give a talk with live Q&A, and that it is developing a communication system combining #languagemodels with new neural decoding strategies to support more natural conversation speed. The company plans to continue human clinical trials in the U.S. and Canada and is engaging people for potential future investigations in the U.K.
28. Tilly Norwood, AI ‘actor’ denounced by actors union, to star in feature film
Particle6 Productions says its #AI-generated “actor” Tilly Norwood will star in a feature-length comedy-drama titled “Misaligned,” escalating a debate over the role of #AI in filmmaking. Creator @Eline Van der Velden described the project as a “hybrid production” pairing traditional film and TV professionals with AI specialists, including AI training and mentorship, and argued AI can support premium narrative work only with substantial human craft and judgment. The actors union SAG-AFTRA condemned the studio for what it called “using stolen performances,” saying Tilly Norwood is not an actor but a computer-generated character with no life experience or emotion, and claiming audiences are not interested in computer-generated content untethered from human experience. Van der Velden told NBC News she is not against AI assisting human creativity and said hybrid methods could include human motion or performance capture alongside real filmmakers to combine human expertise with AI. Set in the “Tillyverse,” a surreal digital world “somewhere up in the Cloud,” the film is framed as a coming-of-age story with “existential AI chaos” that aims to be funny and self-aware while exploring identity, performance, and human fears about #AI.
That’s all for today’s digest for 2026/07/07! We picked, and processed 28 Articles. Stay tuned for tomorrow’s collection of insights and discoveries.
Thanks, Patricia Zougheib and Dr Badawi, for curating the links
See you in the next one! 🚀
