AI News Archive: September 2, 2026 — Part 6
Sourced from 500+ daily AI sources, scored by relevance.
- Researchers build a $7 smartphone clip-on that spots hidden cameras — AI and dynamic LED grid deliver 94% accuracy
This gadget only costs $7 but can help your catch hidden cameras through the use of the companion AI app that installs on your phone. It works by changing the location of the LED light source to compare different reflections and determine if it's a hidden camera or not.
- 11-acre Tampa site on Dale Mabry tied to Tesla's robotaxi expansion
Tesla is expanding its physical footprint in Tampa as the electric vehicle maker builds out its Robotaxi expansion — on a site Waymo wanted first.
Score: 38🌐 MovesSep 2, 2026https://www.bizjournals.com/tampabay/news/2026/09/02/robotaxis-tesla-tampa.html?ana=brss_6150 - Samsung Redefines Laundry With New A-70% and 13 Kg Bespoke AI Washers at IFA 2026
*For illustrative purposes only. Product design and specifications are subject to change. Samsung Electronics today announced that it will showcase two new Bespoke AI Washer models during IFA 2026: a 10 kg model designed to deliver A-70% energy efficiency and a 13 kg large-capacity model engineered to fit within a standard 600 mm depth. As […]
- Horizon Robotics’ Losses Widen as R&D Spending Stays High
Horizon Robotics’ Losses Widen as R&D Spending Stays High Caixin Global
- Voice coding: how developers dictate to Claude Code, Cursor, and Copilot
Explores how developers use voice to command coding assistants like Claude Code, Cursor, and Copilot.
- System helps humans predict when self-driving cars will make mistakes
A new method, called CW-Net, translates the reasoning process of an autonomous vehicle’s AI system into understandable concepts that explain its behavior.
Score: 37🌐 MovesSep 2, 2026https://news.mit.edu/2026/system-helps-humans-predict-when-self-driving-cars-will-make-mistakes-0902 - Real-Time Intelligence with IBM Time Series Models on Confluent
Real-Time Intelligence with IBM Time Series Models on Confluent
- A RAG That Says “Not in This Document” Has to Show Four Kinds of Evidence
Enterprise Document Intelligence [Vol.1 #B3] - A confident wrong answer is a bug. A bare “no answer” with no justification is almost as bad. Each of the four bricks has one piece of evidence to show The post A RAG That Says “Not in This Document” Has to Show Four Kinds of Evidence appeared first on Towards Data Science .
Score: 37🌐 MovesSep 2, 2026https://towardsdatascience.com/a-rag-that-says-not-in-this-document-has-to-show-four-kinds-of-evidence/ - AI boom heats up Bay Area housing market
Here’s another place the AI frenzy is making itself felt: the market for luxury homes
Score: 37🌐 MovesSep 2, 2026https://abcnews.com/Technology/wireStory/ai-boom-heats-bay-area-housing-market-wealthy-136140818 - Chinese internet giants set to reap artificial intelligence profits in 2 to 3 years: UBS
Internet platforms with vast data and large user bases will capture a larger share of artificial intelligence profits in two to three years, even though macro headwinds have temporarily fuelled investor caution over aggressive AI spending by Chinese tech giants, UBS analysts said. Investors have been cautious over a weak macro environment in the second half of the year, and greater AI spending for hardware and infrastructure would drag down short-term profits, but the industry’s power balance...
- State Audit Finds NYC’s AI Oversight Efforts Incomplete
A 2023 audit of New York City’s AI guardrails resulted in three recommendations for improvement. A new report from the state comptroller’s office revealed they have only been partially implemented.
Score: 36🌐 MovesSep 2, 2026https://www.govtech.com/artificial-intelligence/state-audit-finds-nycs-ai-oversight-efforts-incomplete - Aerospace giant's local office secures share of $12.6M unmanned aircraft upgrade deal
The Fairborn branch of an aerospace and defense giant has received a contract modification to a deal worth roughly $2 billion. This update will bring more work to the local employees.
- Shopify is giving its engineers free rein on AI. Here’s why
For decades, e-commerce rewarded the biggest brands that could buy their way to the top of search results. AI is changing that. Shopify COO Jess Hertz shares what Shopify’s data reveals about the agentic shopping revolution in real time, and how the company is building the infrastructure layer that makes products discoverable and purchasable by AI. Plus, why Shopify’s engineers can spend freely on AI tokens while the rest of the business world agonizes over ROI, and what it looks like when a company powering over 14% of online shopping decides to win the future or die trying. This is an abridged transcript of an interview from Rapid Response , hosted by former Fast Company editor-in-chief Robert Safian. From the team behind the Masters of Scale podcast, Rapid Response features candid conversations with today’s top business leaders navigating real-time challenges. Subscribe to Rapid Response wherever you get your podcasts to ensure you never miss an episode. Shopify just posted its, I think, fifth consecutive quarter of growth above 30%. How much of that growth is sort of stealing share to make that happen, like from Amazon or someone else, and how much is about AI? There is a lot of discussion around AI and agentic shopping, but how much business is it really? The way I think about it is really in two pieces. One is our core business, which is incredibly durable, and the second is that AI is expanding our upside. On the core business piece, the best statistic I can give you to think about this is that almost 90% of our quarterly revenue is from merchants who’ve been on the platform for over a year. That growth compounds, and it really shows how strong the business actually is. But on the agentic side, while that agentic GMV [gross merchandise value] is small, it’s definitely growing. We do have early signs that Sidekick is helping merchants sell more, and merchants who are a part of Catalog are really starting to make more money. For listeners who may not be as Shopify-versed, Sidekick is our AI-powered commerce agent that lives within Shopify . What we’re really seeing is that merchants are bringing Sidekick more and more into their business. On the Catalog piece, that’s our product layer for agentic commerce. That’s what makes products understandable, discoverable, and purchasable by AI. What we’re seeing there is that it is increasing conversion. When someone comes from an agentic channel, the conversion rate when an agent is using our catalog data is [double] compared to if it’s just using scraped data. So it’s really the combination of these persistent tailwinds in our base case and, as we enter the agentic era, even more upside. You’re building this sort of infrastructure layer for AI commerce. You have these agentic storefront integrations with ChatGPT and Gemini and Copilot, whatever. How much do you worry that the big AI platforms are going to build their own layer? What I would say is I think we’re past that SaaSpocalypse moment. Shopify has certainly come out the other end of that. The way I really think about that is, complexity is both the challenge and the moat for Shopify. As you mentioned, we’re building all of these products, but when you think about it, a complex world helps Shopify. We’ve always been a unified operating system for commerce. We’ve always acted as this aggregator of different ways to sell, whether that’s an online store, in-person retail, or agentic commerce. And with buyers across different channels, it helps Shopify actually do better because that complexity is hard. It’s hard for merchants to be able to manage that on their own. So I think what’s even more scarce in this AI world is that infrastructure that you’re talking about, right? And building that complete stack for the merchant. And for those merchants who are looking at agentic shopping and this future of having an AI agent between them and a customer, they’ll be relying on Shopify to help them manage that—which is good for you, but might also make them feel a little more beholden, a little more trapped? I don’t think they’ll feel trapped. I think the nice part about Shopify—even if you look at things like the Universal Commerce Protocol, which is the open commerce standard that we developed with Google, which is really the transaction layer of commerce—is that the merchant continues to be the merchant of record. So the idea is not that we’re trying to do anything in terms of disintermediating the merchant. It’s really facilitating for merchants the ability to have that one business and have it work everywhere commerce happens. And so that is really at the core of what our products and partnerships are trying to accomplish for the merchant. I was talking to Uber’s COO, Andrew Macdonald , a few months back. I don’t know if you know Mac, but he ignited this meme about the hidden costs of AI. Yeah. You know, that, “Oh, I’m spending money on these tokens that maybe I’m not getting a return on.” And I’ve heard that your team can still access tokens mostly without restriction. Yes. I think we see a very healthy ROI. Now look, the idea of adoption is never going to be the same thing as impact, so that is definitely true, and we are thinking through ways to measure impact. Head count for Shopify has been flat for more than eight quarters, while revenue is growing 34%. So I think from an ROI, pure financial company perspective, we are doing quite well. I think the other piece to it is that we work carefully in terms of moderating that type of cost. We have a pretty open and flexible view, and we want engineers to be able to use the best models for the best circumstance. Obviously, we have thoughtful speed bumps built in along the way. We’re constantly monitoring it. We’re constantly making sure we’re providing the best tools that we can. But it is not something we take lightly. We are a fairly pragmatic company at bottom. To generate that higher productivity , which is basically what you’re talking about AI doing, doing more revenue with the same number of people, I heard you say that the ideal employee has evolved from being a T-shaped person to an X-shaped person. Can you unpack that for us? It’s another one of my pet theories, so I’ve given you some of my pet theories today. When I joined Shopify, we had this idea of the T-shaped person, right? So you have lots of breadth, but you still have this one vertical that you go deep in, which is your craft, right? For me, that would be a legal craft or a regulatory craft. But what I think we’re starting to see is AI is really pushing toward what I’m calling X-shaped people, right? So you have multiple spikes of expertise, are able to absorb complexity much faster, learn new spaces and be a 7-out-of-10 designer if you want to be, right? Or I try to do my own data work now. So you’re able to move across domains faster. I’ve created this analogy in my head where now you see different slices, right? You have different vectors of expertise, and you’re really moving away from the expectation that you have to have depth in just one area. The other part is really how those different-shaped people actually intersect with each other. I think the idea of team composition and the different constellation of people will only become more and more important as we enter this AI world. It’s actually that team combination that becomes the accelerant to incredible amounts of work that you couldn’t even imagine. And some of that magic is trying to say, “Okay, where are people’s edges? How do they fit together? And how is this sort of shape of employee changing over time?” When you first started describing the X shape, I was thinking, “Oh, this is what we used to call a generalist, maybe a higher-level generalist.” But then when you talk about the teams, they’re generalist specialists because you’re fitting them together in different ways. Exactly. I think a little bit about Tetris, maybe. It’s not that you’re a pure generalist and a pure athlete. It’s that you have the ability to go deep in different areas. You still might have specialized areas of expertise, right? So I still might have a legal background, some craft expertise in certain areas. It’s just that I can learn different areas much faster because I have different tools available to me. So now I not only have a legal or a talent sort of pie or quadrant to me, I’m now deep in commercial, so I’m able to generate these areas of expertise in a much different way than I was five years ago.
- Oracle To Face Earnings Test After Wild Year Riding AI Wave
It was nearly 12 months ago that Oracle stock surged a record 36% on AI optimism. The company has had a rough run since then. The post Oracle To Face Earnings Test After Wild Year Riding AI Wave appeared first on Investor's Business Daily .
Score: 36🌐 MovesSep 2, 2026https://www.investors.com/news/technology/oracle-stock-q1-2027-earnings-ai-openai/ - From Labeling to Deployment: VLM and Agentic AI-Based Autonomous Vision Inspection
From Labeling to Deployment: VLM and Agentic AI-Based Autonomous Vision Inspection
- The agent didn’t leak anything. It just figured something out
Your agent compares a banker’s calendar with the legal team’s and recognizes a pattern: an unannounced transaction is underway. No one told the agent about the deal. It inferred it correctly. Then it adds one line to an executive briefing for a recipient who was not cleared to know about it: “the deal is moving.” Every calendar read was legitimate, and no confidential document was opened. The conclusion is the breach, and no existing permission covers it. Last month I wrote that your next insider threat carries an API token, and that the breach is the sequence of permitted actions, not any one of them. That piece was about what an agent is allowed to do. This one is about what it is allowed to know. The runtime check I argued for there inspects each action before it fires. Here, that check approves every read because each one is permitted. Authorization can travel correctly through every step of the task graph and still miss the synthesized result. Session-based authorization ties access to the current authenticated session. Task-based access control (TBAC) narrows that authority around a specific task; one recent agentic application checks whether the tools an agent requests align with its assigned task. But task scope alone does not automatically answer whether a new conclusion produced from permitted inputs is authorized for a particular recipient. The danger isn’t in any single action. It’s in the join: the agent connects information from authorized sources and produces a conclusion that no single source revealed on its own. That’s aggregation inference. The synthesized result, not the individual inputs, is a new authorization object. It did not exist when the underlying permissions were granted, and no individual permission was written to cover it. What TBAC cannot determine from task scope alone Aggregation inference has predecessors. Intelligence agencies and courts have recognized the mosaic effect for decades: details that appear harmless on their own can reveal sensitive information when combined. Privacy researchers encountered the same limit from another direction. Dwork and Naor examined a formal version of Dalenius’s disclosure-prevention goal: a database should reveal no information about a person that could not be learned without it. They showed that no useful database can meet that standard because a system cannot account for all the outside information a reader may already possess. Access control still has no general answer to either version of the problem. In my recent research, I have been examining aggregation inference as one of three subproblems of authorization propagation in multi-agent systems. An agent can be cleared for every source it touches and still manufacture a conclusion no single clearance covers. The result did not exist until the agent produced it. That work treats the problem as unsolved in the general case. What’s new is that you now employ something that performs the join a thousand times a day, on its own, across everything you let it read — a model whose behavior is not formally specified in advance. It may discover resources dynamically as the workflow unfolds, and the recipient may not know which ones contributed to the conclusion. The shape shows up frequently in the design reviews I sit in. When I threat-model an agent before it ships, the first question is no longer which sources it can read — it’s which sources it can read together. The agents that worry me are never the ones with access to a single sensitive system. They are the ones holding standing read access across two domains whose combination nobody ever reviewed, because each grant looked routine on its own. A January 2026 study by Tianshi Li , run against transcripts from a publicly released interview dataset, shows what individually permissible searches can reveal in combination. The study conducted re-identification tests on 24 interviews in which scientists discussed published work. Web-enabled LLM agents linked six of those transcripts to specific publications, recovering associated authors and, in some cases, uniquely identifying the interviewee. The process bypassed existing safeguards by breaking the re-identification effort into individually benign tasks. Why the floor is not the ceiling One natural response is to classify the conclusion using its source files: take the strictest sensitivity label among what the agent read and apply it to the result. It’s a reasonable instinct, and versions of it are already patented . But the strictest-label approach still cannot solve the problem, and the reason is worth sitting with. Combine the labels of what the agent read, and you learn the floor of sensitivity. You never learn the ceiling. What makes “the deal is moving” sensitive is usually not in any document the agent touched. It is a fact about the world that the agent could not read at all: the board has not announced the transaction yet; an acquisition NDA is in force; a quiet period applies. You can inspect every row the agent saw and never find it because it is not in the data. It is in the world. That is the whole problem. If the property that makes a conclusion dangerous is not in the inputs, then no rule computed from the inputs can catch it. Not the strictest label, not the intersection, not any function of what the agent read. You are trying to classify a fact using only the materials that fail to contain it. That sounds like a dead end. It is actually a direction. If the fact that classifies a conclusion is not in the data, it has to enter the system somewhere a rule can reach, and for the facts anyone can name in advance, there is one place left: the moment a human says what the agent is for. You cannot label the output from the inputs, but a person can label the purpose. The practical starting point is to bind an agent’s authority to a declared purpose. The person who knows what is still secret this quarter can then attach the world-facts that gate that authority: the deal, the embargo and the quiet period. Now the missing fact is in the system, and the machine can enforce policy using it rather than trying to derive it from the inputs. You did not solve the classification. You stopped asking the data to carry a fact it never held. That is the shape of the answer, and it is a long way from shipped. But it tells you which way authority has to point: at the purpose a human declared, not at the files an agent happened to read. So, I will not sell you a fix. Anyone who tells you their product classifies emergent conclusions is selling you the floor and calling it the ceiling. What policies can gate and what requires human judgment What follows isn’t a solution to that classification problem — it’s the lever available today. Cross-domain access rules and combination policies can limit which resources an agent combines and gate delivery based on those inputs. They cannot tell you what the resulting conclusion means. Those controls reduce risk, but they do not solve synthesis authorization in the general case and should not be presented as if they do. In the deal-and-calendars scenario, the immediate step is not to remove access altogether but to assign responsibility for the combination. Someone responsible for the deal’s confidentiality can approve it for a window tied to the matter’s expected duration, re-certify it each quarter while the matter remains open and narrow access when it closes. That turns standing access into an explicit governance decision rather than a default no one remembers granting. Organizations do not need to wait for tooling to name an owner and set the terms. The architectural direction — a design target today, not a shipped control — is to make resource combinations first-class objects of policy: declare which combinations are permitted, evaluate those declarations before a synthesized result is returned, and give agents scoped identities with explicit permissions. Any agent holding standing read access across two sensitive domains at once — people and finance, customers and roadmap, deals and calendars — is not a provisioning ticket. It is a governance decision, and it belongs to someone who knows what is still secret this quarter. Be honest about what this buys you. Gating cross-domain access reduces the number of agents that can perform a dangerous join on their own. It won’t stop every version of this problem. An agent can still read one domain and hand a summary to a person who connects it to something only they know. No access policy will see that final step, because that residual lives in a head, not a document. That exposes the control’s boundary: it can govern what the agent reads but not the conclusion a person ultimately draws from it, a new object that no existing permission covers. The compositions are where the risk lives, and per-resource access control is blind to them by design. If you cannot name the person who owns each agent’s cross-domain access decision, close that gap first.
Score: 36🌐 MovesSep 2, 2026https://www.cio.com/article/4217102/the-agent-didnt-leak-anything-it-just-figured-something-out.html - A Standardized Framework for Assessing High-Risk Life Sciences Research
Recent federal policies establish funding prohibitions and review requirements for certain high-risk life sciences research. The authors advise that federal agencies adopt an objective decisionmaking framework for dangerous gain-of-function research.
- Grok Bot vs. OpenClaw: How I replaced my entire agent stack
Watch now | 🎙️ I’ve been running Grok Bot for several weeks, and I’ve killed every OpenClaw I had. Here are the nine bots doing real work across my inbox, my kids’ school pickups, my PR queue, and my wardrobe
- How AI is playing a growing role in wildfire response
AI is already helping detect and monitor wildfires. Now, researchers are exploring how it could help fire officials decide where to deploy crews across multiple fires. Why it matters: Studies show climate change is contributing to longer, more intense wildfire seasons , forcing officials to make high-stakes decisions about how to prioritize limited resources. AI could help them process information to make those decisions faster. State of play: Jason Fallon, the U.S. Wildland Fire Service's division chief for wildland fire intelligence, told Axios that AI is ingrained in many areas of wildfire management nationally, including monitoring, detection, information dissemination and data transfer. AI-powered cameras, satellite systems and weather-prediction tools are helping agencies in Western states detect wildfires and deploy crews sooner. For satellite data triage, AI can "help us quickly work through a workflow to separate noise from reality," Fallon said. Zoom in: Léonard Boussioux, an information systems professor at the University of Washington Foster School of Business, is part of a team researching how machine learning and optimization could assist fire officials in deciding where to send crews when multiple wildfires are burning. The approach predicts how multiple fires could evolve under different levels of suppression, then uses a mathematical model to recommend how crews could be deployed, according to a research paper that has yet to be peer reviewed. By the numbers: Last year, 77,850 wildfires were recorded across the U.S. — "noticeably higher than the five- and 10-year averages," according to the National Interagency Coordination Center. This year's wildfire season is surpassing the 10-year average for the number of fires and acres burned, according to the National Interagency Fire Center. What they're saying: "What's hard to do is to predict which fire is going to blow up," Boussioux told Axios. "Which fire should we prioritize right now, knowing that we don't necessarily have [the] resources?" Boussioux's team is trying to anticipate how multiple fires could evolve simultaneously. "What we are doing essentially is: We predict two weeks in advance, and we also predict how [multiple fires] will behave" under different levels of suppression, Boussioux said. What we're watching: Fallon said that research like Boussioux's has potential to aid wildfire decision-making, but "I don't think we know yet to what degree." Reality check: AI's role is to augment firefighters, decision-makers and analysts — not replace them, Fallon said. Fire officials remain "in control of directing the strategy and action," he said. AI can reduce cognitive workload and accelerate insights in complex, fast-changing environments, he said, but expertise, experience and risk decisions still need to come from humans. "The machine doesn't have 30 years of experience fighting fire," he said. A model might not know that residents at the end of a road don't have vehicles and need assistance during an evacuation. "It only knows what we can provide to it," Fallon said. Between the lines: AI won't solve the problem of massive, destructive wildfires, Matt Weiner, CEO of Megafire Action, a nonprofit focused on reducing catastrophic wildfire risk, told Axios. "What it can do, and what it's already showing that it can do, is help us prioritize where we can do the most efficient work at every scale," Weiner said. That could mean helping fire officials understand how a new wildfire is likely to behave so they can "direct the right resources for the right fire," he said. AI can provide inaccurate information, potentially putting lives and property at risk, while limited data on rare events can hinder its ability to forecast extreme wildfires, according to the Government Accountability Office. What's next: Researchers are developing more sophisticated tools for modeling complicated wildfire decisions, but their usefulness will depend on the quality and organization of the underlying data. "We already have technology to make things better, but we do not necessarily have data available, or it's not properly managed or organized," Boussioux said. Amy Harder contributed reporting.
- These Russian Mathematicians Taught AI Models How to Talk to Each Other Without Using Words
A startup called Mostik has a wild new approach to combining the capabilities of AI models.
Score: 35🌐 MovesSep 2, 2026https://www.wired.com/story/russian-startup-mostik-ai-models-communication/ - AI is entering the physical world (and your bathroom)
Explores how AI technologies are expanding into everyday physical spaces, including home environments.
Score: 35🌐 MovesSep 2, 2026https://www.superhuman.ai/p/ai-is-entering-the-physical-world-and-your-bathroom - Graph Neural Networks: GCN, MPNN, and GAT, Explained Simply
A visual guide to how graph neural networks work under the hood The post Graph Neural Networks: GCN, MPNN, and GAT, Explained Simply appeared first on Towards Data Science .
Score: 35🌐 MovesSep 2, 2026https://towardsdatascience.com/graph-neural-networks-gcn-mpnn-and-gat-explained-simply/ - This shirt can make you invisible to AI cameras
This shirt can make you invisible to AI cameras YourStory.com
Score: 35🌐 MovesSep 2, 2026https://yourstory.com/ai-story/ai-cameras-hawaiian-shirt-digital-camouflage - AI sounds the death knell for audit fee inflation
Companies’ audit bills increased less than 2% last year, broadly in line with inflation
Score: 35🌐 MovesSep 2, 2026https://www.ft.com/content/80d02797-9a54-47e0-8bc3-a7144354da93?syn-25a6b1a6=1 - Global AI debate lands in the Triangle with OpenAI chief Altman calling technology 'non-negotiable'
The OpenAI CEO told world leaders that AI is driving the largest wave of entrepreneurship and small-business formation in history.
- What Nvidia CEO Jensen Huang said about AI at the G20 in Chapel Hill
Jensen Huang came to Chapel Hill with a warning for world leaders: invest in AI infrastructure now, or fall behind.
- STAT+: How a former ARPA-H director’s startup is tackling AI’s ‘dumb problems’
In this edition of STAT's AI Prognosis: A startup tackling the 'dumb problems' at the intersection of AI and bio, accuracy of AI scribes, and more.
- Countering Model Distillation
Artificial intelligence model distillation (training new models on outputs of more capable ones) could affect U.S. frontier model advantage. New policy tools could augment such technical countermeasures as raising costs or reducing model quality.
- HUMAIN Horizon Ultra AI PC unveiled with Snapdragon X2 Elite and on-device AI
Qualcomm and HUMAIN unveiled Horizon Ultra at LEAP 2026, an AI PC built on the Snapdragon X2 Elite that can run AI tasks locally and switch to the cloud only when it needs extra power.
- OpenMatter adds an AI model gateway and privacy-preserving machine learning
OpenMatter Network Inc. today announced the first significant expansion of its platform since its commercial launch three months ago, adding a developer kit, a gateway for managing access to artificial intelligence models from multiple providers, a rebuilt privacy-preserving machine learning engine and a marketplace for computing across data that never changes hands. The Melbourne, Florida-based […] The post OpenMatter adds an AI model gateway and privacy-preserving machine learning appeared first on SiliconANGLE .
- LIG D&A targets Nordic market with uncrewed vessels
LIG Defense & Aerospace is targeting the Nordic defense market with uncrewed maritime systems designed for surveillance, mine detection and infrastructure protection, the company said Wednesday. At DALO Industry Days 2026, held Aug. 19-20 in Herning, Denmark, the company showcased its Sea Sword unmanned surface vessel, Harbor Underwater Surveillance System and autonomous underwater vehicle for mine detection. It also displayed the AT-1K Raybolt antitank guided missile. The head of the Danish Min
- Irish start-up Setanta plans €3m raise for its AI satellite technology
Setanta has also been accepted into ESA BIC Ireland, a two-year business incubation programme for space-related Irish start-ups. Read more: Irish start-up Setanta plans €3m raise for its AI satellite technology
- Why the 'Brain' of China's AI Runs on a Grassland: Inside Alibaba's Ulanqab Data Center
Alibaba's Ulanqab data center shows how cheap renewable power and low latency—about 4 milliseconds to Beijing—are making a remote Inner Mongolia city into a center of gravity for China's AI compute.
- Scientists Warn About a Creepy New AI Side Effect on Humans
New research suggests the more humans interact with AI chatbots, the more they act and talk like robots themselves.
Score: 33🌐 MovesSep 2, 2026https://www.inc.com/jessica-stillman/scientists-warn-about-a-creepy-new-ai-side-effect-on-humans/91398406 - JD Vance talks end times, Antichrist and AI on faith podcast
The vice president, who has been an advocate of building data centers to support artificial intelligence, told podcaster Bryce Crawford that some uses of AI are “satanic.”
- Almost half the time spent on AI is on fixing its output, BambooHR says
U.S. employees still remain largely positive about the technology at work, though younger workers are wary that it could limit their growth potential.
Score: 33🌐 MovesSep 2, 2026https://www.hrdive.com/news/almost-half-the-time-workers-spend-on-ai-is-spent-fixing-its-output/829404/ - Breakingviews - COMMENTARY: Financiers are set to turn Nvidia into an AI baron
Breakingviews - COMMENTARY: Financiers are set to turn Nvidia into an AI baron reuters.com
Score: 33🌐 MovesSep 2, 2026https://www.reuters.com/commentary/breakingviews/financiers-are-set-turn-nvidia-into-an-ai-baron-2026-09-02/ - Texas and Florida Step Back from ALPRs
Within the last few days, two important state actions have dealt a big blow to automated license plate reader (ALPR) networks. This is just the latest proof of the growing tide of public opposition to mass surveillance. After years of successful grassroots battles to pull these cameras from local streets, bipartisan momentum is sweeping the country. On August 28, Texas Governor Greg Abbott banned state agencies from spending public funds on Flock cameras . The order dropped just as The Texas Tribune prepared to publish an investigation revealing that a state agency had quietly funneled at least $30 million into building a sprawling surveillance network. Then on August 31, the Florida Department of Transportation (FDOT) issued a memo , announced by Governor Ron DeSantis, ordering the removal of all ALPRs from the right-of-way on state highways within 30 days. The order revokes all previously approved permits to install ALPRs, and bars transportation officials from issuing future permits. FDOT officials stated that “the recent exponential increase in deployments along our roadways, coupled with concerning reports of misuse, data privacy concerns, and surveillance schemes merit immediate action to preserve Floridians’ sovereignty and quality of life.” FDOT’s action has been followed by a surge of local governments in Florida canceling or pausing their vendor contracts . Much more work remains. Many ALPRs in Florida are not on state highways, but sit on city streets, county roads, residential driveways, and shopping center parking lots—and FDOT's order doesn't touch any of them. Likewise, the Texas directive leaves local agencies free to use city, county, federal, and private funds to install cameras. This week’s good news follows years of pushback from local advocates that has seen dozens of cities sever ties with surveillance companies. According to some metrics , during the last 30 days, an average of three localities per day has halted contracts with Flock. Other advocates have been resisting ALPRs in statehouses and court houses , and by blowing the whistle with investigative activism. The moves in Florida and Texas also illustrate the power that the executive branch can wield to curtail mass surveillance with almost immediate results. We hope that the California Governor Gavin Newsom and the California Department of Transportation will take notice and initiate steps to curb this technology, starting with removing the ALPRs that U.S. Border Patrol and the Drug Enforcement Administration have installed on California highways. EFF’s position remains : ALPR mass surveillance – the indiscriminate, continuous collection and retention of location data on every driver, regardless of suspicion – should not exist. This past week’s actions in Texas and Florida are good steps forward, but we are still far from the finish line. We will continue working alongside community groups to keep cameras off local streets, while urging judges and state lawmakers to impose enforceable restraints on this warrantless mass surveillance.
- How Coinbase used Code Connect to guide agents and shrink token costs
The Coinbase Design System team put Code Connect to the test against coding agents—and found that it boosted design system adherence while cutting token costs by an average of 22.5%.
Score: 33🌐 MovesSep 2, 2026https://www.figma.com/blog/how-coinbase-used-code-connect-to-shrink-token-costs/ - Meet Switchyard: A Rust Proxy and Library That Routes and Translates LLM Traffic Across OpenAI and Anthropic APIs
Meet Switchyard: A Rust Proxy and Library That Routes and Translates LLM Traffic Across OpenAI and Anthropic APIs MarkTechPost
- Pangram Has Emerged as the Gold Standard of AI Detection. Should You Trust It?
Meet the AI police who can make or break careers—in publishing and beyond.
Score: 33🌐 MovesSep 2, 2026https://www.wired.com/story/pangram-has-emerged-as-the-gold-standard-of-ai-detection/ - Next-gen AI networks may hinge on the telephone switchboard's return
Photonics startups like iPronics are raking in hundreds of millions in funding to make optical circuit switches faster, denser, and cheaper
- NASA-linked, MIT-trained founders’ nSWX raises US$2M for AI chip packaging
The race to build larger AI models has created a less glamorous but increasingly urgent problem: how to move heat and power through ever-denser chip packages without driving up cost, energy use and complexity. A Malaysian startup now wants to solve part of that problem at the material boundary level. Kuala Lumpur-based nanoSkunkWorkX (nSWX) has […] The post NASA-linked, MIT-trained founders’ nSWX raises US$2M for AI chip packaging appeared first on e27 .
Score: 32💰 MoneySep 2, 2026https://e27.co/nasa-linked-mit-trained-founders-nswx-raises-us2m-for-ai-chip-packaging-20260902/ - Podcast: We Spoke to an Amazon Worker Destroying Books for AI
A follow up to the Amazon destroying books for AI story, why a bunch of names keep appearing in AI-generated papers, and ICE's latest spending spree.
Score: 32🌐 MovesSep 2, 2026https://www.404media.co/podcast-we-spoke-to-an-amazon-worker-destroying-books-for-ai/ - Rethinking cybersecurity operations in the age of artificial intelligence
[The content of this article has been produced by our advertising partner.] Cybersecurity is evolving from a manual operation to one driven by artificial intelligence (AI). The volume, speed and sophistication of modern cyber threats now exceed the capabilities of conventional processes to handle them alone. AI is reshaping the operating model by automating repetitive, high-volume work while improving accuracy and efficiency across the security life cycle. By accelerating threat detection...
- Berkshire CEO Abel says AI to help power growth
Berkshire CEO Abel says AI to help power growth reuters.com
- Americans Love AI. They Don't Want the Data Centers That Power It.
Americans Love AI. They Don't Want the Data Centers That Power It. Business Insider
Score: 32🌐 MovesSep 2, 2026https://www.businessinsider.com/americans-love-ai-they-dont-want-the-data-centers-2026-9 - Plastic surgery patients asking for ‘AI faces’
Plastic surgery patients asking for ‘AI faces’ The Telegraph
Score: 32🌐 MovesSep 2, 2026https://www.telegraph.co.uk/news/2026/09/02/plastic-surgery-patients-asking-ai-faces/ - Turn detection in voice agents: silence timeouts vs VAD vs semantic models
Compare silence timeouts, VAD, and semantic turn detection for voice agents, then measure false interruptions and tune Flux STT.
Score: 32🌐 MovesSep 2, 2026https://deepgram.com/learn/turn-detection-voice-agents-silence-vad-semantic-models - Unitree employees reportedly describe founder-led management and high internal pressure
Current and former Unitree employees have reportedly described a tightly centralized management style and a demanding work environment at the Chinese robotics company. Accounts circulating online alleged that expenses above RMB100 required approval from CEO Wang Xingxing and that he was closely involved in product details. The reports also described strict confidentiality requirements and demanding […]