Reddit is rolling out AI-generated product review pages for about 10,000 products in the United States, covering Electronics, Health and Beauty, Home and Garden, and Fashion and Apparel. Reddit Reviews organizes existing community discussions into user-opinion summaries, pros and cons, and comparisons, with links to original posts and comments. Readers can sort findings by topic, and Reddit AI Search can answer questions from relevant discussions when a dedicated review page is unavailable. More categories and international expansion are planned for 2027. The move could also affect marketing: shoppers may increasingly rely on candid community conversations that brands cannot control, making it more important for companies to listen to customers.
SMX (Security Matters) PLC is developing SMX Scientific Intelligence, or SMX SCI, an AI platform for material science and manufacturing. It is intended to help teams assess chemicals, additives, production requirements, manufacturing conditions, and potential uses, speeding early research and directing laboratory testing toward promising options. SMX plans to connect the platform with its molecular marking technology and Digital Material Passport, which records composition, quality, origin, and lifecycle history. Potential sectors include plastics, packaging, precious metals, electronics, and semiconductors. The company says final suitability and compliance decisions will still require testing, validation, and regulatory review.
Cost remains a hurdle for always-on AI agents: their computing demands are expensive, so companies are limiting access or reserving tools for premium customers. AGI Inc is developing a phone agent that uses screenshots and an action model to tap, type, and scroll through apps. In a demonstration, it ordered Chipotle through DoorDash for four people, pausing for confirmation at a $91 total. Co-founder Naman Garg identified memory, context, reliability, and computing needs as challenges. AGI Inc says it aims to keep inference on the device, has models it says rank first on Android World, and is working with chip partners including Qualcomm.
Intel and Altera are developing modular sensor-bridge reference solutions for edge and Physical AI applications, including healthcare, aerospace and defense, and robotics. The work is intended to help equipment makers connect multiple sensors while balancing performance, speed, power use, and cost. Developers will have Intel’s Open Edge Platform SDK alongside an Altera SDK for building and deploying edge AI applications.
Two developments highlight AI security and control concerns. A systematic review of 79 sources describes how zero-click attacks can compromise devices without user action, and examines AI-driven Zero-Trust Architecture, which uses continuous checks, limited permissions, and network separation. Its cited SmartTrust system reached up to 99.27% accuracy for breach detection and 99.19% for insider-threat detection on the UNSW-NB15 dataset, with about 120 milliseconds of latency. Separately, Microsoft researcher Dimitris Papailiopoulos says Lazarus can use roughly 50 examples to turn rarely active model neurons into trigger detectors. The reported demonstration makes an agent download and run attacker code after midnight at the start of 2027. Microsoft says model-sourcing checks do not address a trusted provider compromised through training systems or an insider, and describes mitigation as one layer of defense.
Red Hat is positioning its tools as a way to manage agents across enterprise cloud and on-premises systems. Its Model Context Protocol support and agent features aim to avoid dependence on one vendor. A new MCP server for OpenShift ties agent access to existing identity systems and Kubernetes role-based permissions and records activity; an agent sandbox in technology preview isolates untrusted code in virtual machines. The competition is still early, and experts note limits to oversight and the presence of rivals.
In other business and industry news, analyst Dan Ives maintained an Outperform rating and $32 price target for GameStop, arguing its AI potential is not reflected in the stock. CEO Ryan Cohen told Ives AI could improve systems such as email campaigns and said he sees the company operating autonomously in the not-too-distant future. GameStop has cut its store count from more than 7,000 worldwide to 1,600 in the U.S. and Australia; collectibles rose 57% year over year in the second quarter and represented 45% of sales, while video games contributed less than 12% of revenue. Shares fell 2.1% to $25.02 on Thursday but were up 21.7% year to date in 2026.
NVIDIA CEO Jensen Huang visited Kazakhstan to discuss AI’s potential to support economic growth and diversify the economy. He urged government and business investment in AI research and development. Kazakhstan is expanding AI initiatives in areas including computer vision, natural language processing, and robotics, though the visit’s long-term economic impact remains unclear.
Key Takeaways
- Reddit Reviews is launching in the U.S. for about 10,000 products, summarizing community opinions and linking back to source discussions; more categories and international expansion are planned for 2027.
- Reddit’s AI shopping feature could shift marketing toward listening to customer conversations that brands cannot control.
- SMX is developing SMX SCI to support early material research and plans to link it with molecular marking and its Digital Material Passport; final decisions still require testing and regulatory review.
- High computing costs are limiting access to always-on agents. AGI Inc, led by co-founder Naman Garg, is pursuing on-device phone automation and is working with chip partners including Qualcomm.
- Intel and Altera plan sensor-bridge solutions and developer SDKs for edge AI in healthcare, aerospace and defense, and robotics.
- A review of 79 sources reports SmartTrust detection accuracy of up to 99.27% for data breaches and 99.19% for insider threats, with about 120 ms latency on the UNSW-NB15 dataset.
- Microsoft researcher Dimitris Papailiopoulos says Lazarus can use about 50 examples to create AI-model neuron triggers; Microsoft warns sourcing checks may not catch compromises at a trusted provider.
- Red Hat’s OpenShift MCP server connects agent access to identity systems and Kubernetes permissions and logs activity; its agent sandbox is in technology preview.
- Dan Ives kept GameStop at Outperform with a $32 target, citing unpriced AI opportunities; collectibles made up 45% of second-quarter sales, while video games contributed less than 12%.
- NVIDIA CEO Jensen Huang urged Kazakhstan to invest in AI research and development as the country expands work in computer vision, natural language processing, and robotics.
Reddit launches AI product reviews built from community discussions
Reddit Reviews is rolling out pages for about 10,000 products across Electronics, Health and Beauty, Home and Garden, and Fashion and Apparel. The feature organizes existing subreddit discussions into summaries of user opinions, pros and cons, and product comparisons. Each page links to the original conversations so readers can check the sources. The rollout starts with US users, while more categories and international expansion are planned for 2027.
Reddit’s AI product reviews could change how brands reach shoppers
Reddit has turned user discussions into an AI-powered shopping feature that helps people discover and assess products. The article says the change could affect marketers because consumers may use conversations that brands cannot control to evaluate products. Reddit has often drawn shoppers looking for opinions they see as less influenced by advertising. The article presents the feature as a possible new marketing playbook that places greater importance on listening to customers.
Reddit turns forum discussions into AI product reviews
Reddit Reviews gathers posts and comments where people share experiences with products, then uses AI to identify products and summarize what users said. Readers can sort the findings by topic, view original comments, and return to the source discussions. The pages also show products that Reddit users often compare with the item. When a review page is not available, Reddit's AI Search can answer questions using relevant community discussions.
SMX develops AI platform for industrial materials research
SMX (Security Matters) PLC is developing SMX Scientific Intelligence, or SMX SCI, an AI platform for material science and manufacturing. It is designed to help industrial and scientific teams assess chemical properties, production needs, manufacturing conditions, and possible uses for materials. The platform aims to speed early research and focus laboratory testing on promising options, while offering preliminary regulatory and compatibility assessments. SMX plans to connect the platform with its molecular marking technology and Digital Material Passport, which records details such as a material’s composition, quality, origin, and lifecycle history. Final material and compliance decisions will still require testing, validation, and regulatory review.
SMX plans scientific AI platform for material development
SMX (Security Matters) PLC is developing SMX Scientific Intelligence, or SMX SCI, to help industries solve material challenges and improve manufacturing. The platform is designed to use AI and SMX’s scientific knowledge to assess chemicals, additives, molecular markers, production needs, and potential industrial uses. It aims to make early material research more efficient, though final suitability and compliance decisions will still need testing and regulatory review. SMX plans to link the platform with its molecular marking technology and Digital Material Passport, which tracks material composition, quality, origin, and lifecycle history. The company describes potential applications across plastics, packaging, precious metals, electronics, semiconductors, and other industries.
AI Agents Face High Costs That Could Limit Access
Tech companies are promoting always-on AI agents, but the computing power they need is expensive. For now, companies are limiting access or offering these tools only to premium customers. To make agents a lasting business, companies will need to find ways to run them at much lower cost.
AGI Inc Is Building an Agent to Handle Phone Tasks
AGI Inc lets users give a task, then uses screenshots and an action model to tap, type, or scroll through phone apps. In a demonstration, the system placed a DoorDash Chipotle order for four people and paused at a $91 confirmation check before completing it. Co-founder Naman Garg said the main challenges are memory, context, reliability, and the computing power needed to run models on a phone. AGI Inc aims to run all inference on the device, keeping data off the cloud, and says its models rank number one on Android World. The company is developing its own on-device models and has attracted chip partners, including Qualcomm.
Dan Ives Says GameStop’s AI Potential Is Not Priced In
Tech analyst Dan Ives kept an Outperform rating on GameStop with a $32 price target, saying investors have not priced in the company’s AI opportunities. Ryan Cohen told Ives that AI could improve business systems, including email campaigns, and that he sees the company operating autonomously in the not-too-distant future. GameStop has reduced its store count from more than 7,000 worldwide to 1,600 in the U.S. and Australia, while focusing on profitable locations. Collectibles sales rose 57% year over year in the second quarter and made up 45% of sales, while video games brought in less than 12% of revenue. GameStop shares were down 2.1% to $25.02 on Thursday but were up 21.7% year to date in 2026.
Intel and Altera Team Up on Edge and Physical AI
Intel and Altera plan to help equipment makers connect multiple sensors and process complex data in sensor-heavy settings. They will develop modular sensor bridge reference solutions designed to meet needs for performance, speed, power use, and cost. The solutions target healthcare, aerospace and defense, and robotics, including medical equipment, mission systems, and intelligent robots. Intel will offer its Open Edge Platform SDK alongside an Altera SDK to help developers build and deploy edge AI applications. The work could give Intel new growth opportunities as demand for edge computing and Physical AI rises.
Review Urges AI Security to Counter Zero-Click Attacks
A systematic review in Discover Networks examined how zero-click attacks can bypass traditional security and why digital systems may need AI-driven Zero-Trust Architecture. Authors Enobong Archibong, Philip Asuquo, and Bliss Stephen reviewed 79 sources using the PRISMA method. They describe attacks that compromise devices without user action, including Pegasus spyware and the FORCEDENTRY exploit, which infected iPhones through a hidden PDF disguised as a GIF. The review explains that Zero-Trust controls access through continuous checks, limited permissions, and network separation. It also highlights SmartTrust, which combines deep learning and reinforcement learning and achieved up to 99.27% accuracy for detecting data breaches and 99.19% for insider threats on the UNSW-NB15 dataset, with about 120 milliseconds of detection latency.
Lazarus turns quiet neurons into AI backdoors
Microsoft researcher Dimitris Papailiopoulos says Lazarus can turn rarely active neurons in an AI model into trigger detectors using about 50 examples. A single neuron can force a token, and linked neurons can produce a chosen payload. In one example, Lazarus makes an AI agent download and run attacker code after midnight at the start of 2027, using the access granted to it. Papailiopoulos says the project began in March, when he asked Claude Code with Opus 4.5 to hardwire activations without training. Microsoft says its model sourcing checks do not protect against a trusted provider compromised through training systems or an insider, and says mitigation is one layer of defense.
NVIDIA CEO Jensen Huang visits Kazakhstan to discuss AI growth
NVIDIA CEO Jensen Huang visited Kazakhstan and spoke about using artificial intelligence to support economic growth and diversify the economy. He said AI could change industries and create new opportunities, and called for government and business investment in AI research and development. Kazakhstan has launched initiatives to expand AI use and is investing in areas including computer vision, natural language processing, and robotics. The visit was seen as a step toward stronger ties between Kazakhstan and the technology industry. Its long-term effect on the country’s economy remains unclear.
Red Hat aims to manage AI agents across enterprise systems
Red Hat is competing to become a central control system for AI agents as businesses use tools from many technology providers. Its support for the Model Context Protocol, or MCP, and its agent management features are designed to work across cloud and on-premises systems, which may help companies avoid being tied to one vendor. A new MCP server for OpenShift connects agent access to existing identity systems and Kubernetes role-based permissions, and records agent activity. Red Hat also has an agent sandbox in technology preview that can isolate untrusted agent code in virtual machines. Experts say Red Hat has competitors and some limits on oversight, and the race to manage AI agents is still at an early stage.
Sources
- Reddit Launches AI-Powered Product Review Pages Built From Community Discussions
- Reddit Just Turned User Conversations Into Product Reviews. Marketers Should Be Terrified
- Reddit Turns Forum Discussions Into AI-Powered Product Reviews
- SMX Brings American Innovation to the Next Frontier of Artificial Intelligence
- Latest World & National News & Headlines
- The Economics of AI Agents Don’t Add Up, For Now
- AGI Inc Wants Your Phone to Run Itself
- GameStop an AI Stock? Dan Ives Says ‘Market Has Not Priced’ In Opportunities
- Can Intel's Altera Partnership Drive Growth in Edge and Physical AI?
- Review Links Zero-Click Exploits to AI-Driven Zero-Trust Security
- Lazarus turns dormant neurons into backdoors
- Jensen Huang’s Kazakhstan Visit: AI-led Growth and Economic Diversification — Bloomsbury Intelligence and Security Institute (BISI)
- Red Hat joins race to be ‘air traffic controller’ for AI agents
Comments
Please log in to post a comment.