AI News Archive: July 15, 2026 — Part 7
Sourced from 500+ daily AI sources, scored by relevance.
- Sarvam, BharatGen offer AI cheaper than DeepSeek. Can they sustain it?
With Indian firms far behind global rivals, experts cautioned that these may not have the kind of global pricing impact that China did with DeepSeek
Score: 46🌐 MovesJul 15, 2026https://www.livemint.com/ai/sarvam-bharatgen-ai-deepseek-open-ai-11784007352928.html - IITPSA announces major ICT skills conference to deep dive into AI’s impact on jobs
The conference will delve into the findings of this year’s IITPSA Skills Survey, which will analyse the impact of AI on ICT jobs in SA.
- Salesforce's Agentforce isn't winning over clients, KeyBanc analysts claim
Investment bank cites messy customer data and a product that 'just isn't there'; Salesforce counters by saying it the fastest-growing product in its history
- AWS adds AI-assisted product listing service to its Marketplace portfolio
Product listings in AWS Marketplace gained new AI-based features last month in anticipation of continued growth in the use of enterprise agents. Amazon Web Services Inc. unveiled AI-assisted product listings in Product Assistant chat, a feature that helps independent software vendors and consulting partners develop comprehensive product listings for AWS Marketplace using existing digital assets. […] The post AWS adds AI-assisted product listing service to its Marketplace portfolio appeared first on SiliconANGLE .
Score: 45🌐 MovesJul 15, 2026https://siliconangle.com/2026/07/15/aws-product-listing-service-enterprise-agents-ai-awsmarketplaceseries/ - AI Can Summarize Employee Feedback. A New Benchmark Shows It Doesn’t Always Understand It.
AI Can Summarize Employee Feedback. A New Benchmark Shows It Doesn’t Always Understand It. Toronto Star
- Octave Launches CoLabs: A Customer Innovation Program to Scale AI Across Industrial Operations
Octave Launches CoLabs: A Customer Innovation Program to Scale AI Across Industrial Operations Toronto Star
- The hidden cost of AI: Why finOps is becoming critical for AI-driven organisations
By Naman Aggarwal, CGMO at CloudKeeper For most organisations, the conversation around AI starts with possibilities. Faster development. Smarter customer experiences. Better productivity. New products. New revenue streams. The conversation […] The post The hidden cost of AI: Why finOps is becoming critical for AI-driven organisations appeared first on Express Computer .
- What problems would an AI speaker from OpenAI actually solve?
OpenAI’s first device will be a screenless home AI system that can play music, control appliances, and respond to messages and questions, according to Bloomberg . I can’t help but ask what makes this device different from Apple’s HomePod with SiriAI? The OpenAI product is intended to be the first of a family of solutions and is expected to use the recently-introduced GPT-Live large language model (LLM). The latter is an advanced model capable of processing information swiftly and of providing natural responses to conversations. A potential gold mine for hackers This expertise might help it deliver more accurate responses to requests, though the information could also become a gold mine for data brokers, hackers, and advertisers if there turns out to be any way they can get their hands on it. It’s not yet known how — or even if — OpenAI proposes protecting user privacy within its systems. The device, which is still under development, is explained as being a home companion that also includes a built-in camera and sensors so it can gather contextual information about where you are, becoming an expert on you and your needs. It also features autonomous mechanical elements that physically shift on their own, intended to give the product a “personality,” rather than being a boring black box. Can we live without this? While OpenAI’s development is not yet complete and things could change before it reaches market, based on Bloomberg’s report I’m not terribly clear how unique it is going to be. After all, as LLM support is introduced in existing smart speaker systems from Apple, or even Amazon, what unique features does this device bring that consumers can’t live without? More particularly, what problems does it solve and why does it exist? Manufacturing economics What’s also unclear is how far along OpenAI is on the road to mass manufacturing the device. Apple’s recent lawsuit against OpenAI confirmed the challenger is speaking with Apple’s own manufacturing partners as well as hiring hundreds of Apple engineers . Despite the talent war, I consider it unlikely OpenAI will be able to lock in the kinds of manufacturing deals it needs to bring the product to market at an acceptable price . That suggests either that its inaugural “home companion” will seem incredibly expensive (as so many of the products Jony Ive has designed since leaving Apple seem to be), or that OpenAI will sell these things at a subsidy. Bloomberg suggests the systems will cost $200 to $300. That seems low given the current component market, expected design quality and the technology used if the plan is to make something good. And it leaves me wondering how deeply investors will underwrite hardware sales, given the eye-watering losses the company is already making . Apple’s trade secrets case Given ongoing speculation that Apple plans something similar in the form of a hybrid HomePod/iPad equipped with AI and limited mobility, the recent lawsuit strongly suggests Apple feels some of OpenAI’s plans cross the line into using proprietary technologies and ideas Cupertino has spent years pursuing. Apple’s lawsuit seems to bring much more meaningful evidence than just an argument concerning product design. Betting the farm on Ive We also don’t know the extent to which consumers will be open to semi-sentient AI devices lurking in their lives. While Apple can lean into its loyal customer base and broaden its offering with rock-solid promises concerning user privacy, OpenAI has less to bring to the launch party. That means it is attempting to pivot millions who use its services into investing in its hardware. It presumably hopes that it will be able to drive that transition by using the design involvement of acclaimed Apple designer Jony Ive as a form of magic talisman. The challenge is that while Ive is a big name in Apple history, Apple users are extremely loyal and may react against the involvement of their favorite designer. It’s like finding out someone you thought was on your team actually supported someone else. It will be different outside Apple, where less loyal cohorts might see the product introduction as a chance to put a design from Ive through its paces without signing up to a Mac, iPhone, iPad, or HomePod. For the rest of us, the question will be whether OpenAI’s Ive-designed product channels the successful design ethic of the iMac, or that of the far less successful hockey puck mouse. Like (timely World Cup klaxon) France against Spain, OpenAI’s investors have to hope the best version of Ive’s design principles show up, because their risked fortunes potentially depend on it. You can follow me on social media! Join me on BlueSky , LinkedIn , Mastodon and subscribe to The Core .
Score: 45🌐 MovesJul 15, 2026https://www.computerworld.com/article/4197338/what-problems-would-an-ai-speaker-from-openai-actually-solve.html - 5 ways for CIOs to avoid AI bill shock
Gen AI spending is moving beyond the familiar software model of seats, licenses, and pilots. As AI shifts from copilots to embedded workflows and autonomous agents, one user request can trigger multiple model calls, retrieval steps, retries, orchestration layers, and infrastructure events. A tool that looks affordable in pilot may behave very differently once connected to production systems or allowed to act with less human supervision. According to Michael Corrigan, CIO of World Insurance Associates, AI introduces a fundamentally different cost model — one that’s usage driven, non-linear, and tightly coupled to business activity. “Success requires shifting from traditional IT budgeting to FinOps-style discipline where consumption, value, and governance are actively managed in real time,” he says. Here are five ways CIOs can build that discipline before AI costs spiral. Forecast AI by workflow, not by user At World, a top 25 insurance broker with about 3,000 employees across roughly 300 locations, AI use falls into three broad categories, Corrigan says. One is broad tools, such as copilots. Another is embedded AI inside SaaS platforms. And the third is bespoke AI built around specific workflows and manual processes. “The bespoke is the area that’s growing the most right now,” he says. “And that’s where the model, from a cost perspective, has really been shifting from a license seat cost to a token consumption or token burn cost, or even a hybrid.” width="1240" height="827" sizes="auto, (max-width: 1240px) 100vw, 1240px"> Michael Corrigan, CIO, World Insurance Associates WIA Seat-based pricing is relatively easy to forecast whereas consumption-based AI isn’t. Costs may depend on prompt complexity, output length, model choice, workflow design, and whether the system calls a model once or many times in the background. World tries to manage that uncertainty by defining the business problem, success criteria, and expected operational improvement upfront. Pilots help estimate consumption before scaling, but Corrigan says they don’t remove the ambiguity. “We’ll try our best in the pilot to understand what the consumption rate is, what the token burn rate is,” he says. But once a consumption-based workflow goes into production, he adds, an estimate is put into place. That estimate is informed, but still rough. Elmer Morales, founder and CEO of koder.com, an agentic AI coding startup, says CIOs should think less about headcount and more about workflow mechanics . Agentic AI costs are driven by the number of decisions an agent makes, how often it retrieves external data, how much context it carries, and how many systems it touches. “CIOs should start by mapping workflows, not necessarily users,” he says. “The relevant variable isn’t going to be the headcount but how many decisions an agent makes per task.” Model the failure path, not just the happy path Pilots can mislead because they often test the cleanest version of an AI workflow. Morales says many enterprises model agentic AI costs around the happy path: the user gives a clear prompt, the system understands the request, the agent completes the task, and the process ends. Production is messier. “They generally don’t model for situations where the agent is going to need to go back and check its work and redo things,” Morales says. “A lot of times, agents are wrong, either because they hallucinate or they understood the problem incorrectly.” In an agentic workflow, the system may check its work, call another tool, retrieve more data, or redo a step. While that may improve quality, it also adds cost. width="1240" height="827" sizes="auto, (max-width: 1240px) 100vw, 1240px"> Elmer Morales, founder and CEO, koder.com koder.com The difference between copilots and agents is central. A copilot interaction is often one prompt and one response. An agentic workflow may involve agents moving through a decision tree, executing tasks in sequence or in parallel, and calling sub-agents or external systems along the way. “By the time it’s achieved the original goal, the agent might have made 50 or 100 model calls, compared with a single call for a traditional copilot prompt,” Morales says. That’s why CIOs should require teams to model the failure path before production, like how many retries are allowed, how much context is resent, which tools can be called, when a human should intervene, and what happens when the agent can’t complete the task. Build cost controls into the architecture Traditional FinOps practices still matter, but AI requires more than retrospective dashboards and chargebacks. According to Pavan Madduri, senior cloud platform engineer at industrial supply company W.W. Grainger, looking backward at usage data, as traditional FinOps often does, can be too late. Costs are shaped by prompt design, model selection, agent behavior, orchestration choices, and runtime loops. “Dashboards or chargebacks, those are historical accounting,” he says. “The money’s already gone.” For AI, he argues, cost controls need to be embedded into the architecture. That includes hard token caps, retry-depth limits, maximum runtime limits, workload prioritization, background-job throttling, and cluster-level controls that prevent runaway consumption. “The real FinOps means you need to have the cost constraints embedded into your architecture framework,” Madduri says. Those controls also extend to infrastructure. Expensive GPUs may sit warm between jobs because systems need capacity available when inference demand arrives. Teams may pass huge schemas, databases, or thousands of lines of code into frontier models when a smaller or more focused prompt would do. width="1240" height="828" sizes="auto, (max-width: 1240px) 100vw, 1240px"> Pavan Madduri, senior cloud platform engineer, W.W. Grainger W.W. Grainger Enterprises should also adopt event-driven autoscaling, Madduri says. “Use tools like KEDA to scale GPU nodes down to zero the moment inference demand drops, so teams only pay for the windows when the silicon is actively crunching tokens.” Corrigan says World uses rate limits, spend limits, alerts, and approval gateways for consumption-based tools. When users approach token consumption limits, automated alerts allow IT and the business to review whether the continued spend is justified. “If it’s not meeting the success criteria we expected, you have to have the control in place to say we’re going to move on or kill that process,” Corrigan says. Route work to the right model CIOs can also reduce AI bill shock by avoiding a default assumption that every task requires the most powerful model available. While some tasks need advanced reasoning, many others don’t. A simple support ticket, log-parsing task, or structured database transaction may be handled by a smaller or cheaper model. A complex architecture decision, legal analysis, or multi-step reasoning task may justify a more powerful one. “Choosing the right model for the right prompt and right question — that’s where you leverage the maximum from that model, and you can decrease the costing,” Madduri says. “If you default every single call to a frontier model, that’s architectural laziness.” Morales makes a similar point. Not every step in an agentic workflow requires a top-of-the-line model. Model routing, he says, is the discipline of determining the best model for the task, and providing the relevant context when the model needs it. According to Jim Olsen, CTO of enterprise software company ModelOp, CIOs should use the least expensive model that can accomplish the business goal. Using the biggest model for everything is easier, but expensive. “It’s like hiring the most expensive engineer to change a few colors in a website’s CSS, or visual styling,” he says. “You wouldn’t do that. You use the appropriate tools for the task.” Tie consumption to business value For Olsen, the deeper enterprise problem is AI value shock, not just bill shock. Spending $200,000 in a quarter on AI is justified if it produces $2 million in business value. The problem is spending heavily on use cases that don’t generate a meaningful return. “Are you actually getting that return on investment, or are you just blowing tokens for something that’s not delivering the value to your business?” Olsen asks. Tracking token usage by user or department may show who consumed AI, but not whether the consumption mattered. width="1240" height="827" sizes="auto, (max-width: 1240px) 100vw, 1240px"> Jim Olsen, CTO, ModelOp ModelOp For most enterprise AI systems, Olsen says costs should be tied back to business use cases. A model may be used for HR document search, customer support, code review, problem resolution, or other functions. Each use case may draw on the same underlying models or agents, but the business value can be very different. That’s why he argues that companies need an AI inventory, a record of which business workflows use which models, agents, providers, workflows, and systems. Without that inventory, enterprises can’t connect consumption to value. Corrigan takes a similar approach from a governance perspective. At World, new AI ideas go through an intake process. Business users propose improvements, and IT, finance, operations, sales, and business stakeholders evaluate, prioritize, and monitor them from pilot through production. That may be where the next stage of AI FinOps is heading, toward a clearer understanding of which AI consumption deserves to scale, not just to lower bills. So the question, as Olsen puts it, isn’t whether someone used a million tokens. It’s what are they using them for.
Score: 45🌐 MovesJul 15, 2026https://www.cio.com/article/4190605/5-ways-for-cios-to-avoid-ai-bill-shock.html - I let ChatGPT Work and Claude Cowork loose on my files - only one made me nervous
How does ChatGPT Work compare with Claude Cowork for desktop automation? My testing reveals similar results, similar strengths, and one major reason Claude currently feels considerably safer right now.
- Relax, Apple, OpenAI and its rumored AI smart speaker plans are no threat to you, Siri, HomePods, robots, or any other part of your business
Fresh rumors point to OpenAI working on a human-like, screenless AI-powered smart speaker. If true, it sounds terrible.
- Here’s the Truth About Whether Meta’s NameTag Face Recognition Tech ‘Exists’
Since WIRED reported on Meta’s NameTag face recognition system, company executives have made confusing and conflicting remarks about its very existence.
Score: 45🌐 MovesJul 15, 2026https://www.wired.com/story/heres-the-truth-about-whether-metas-nametag-face-recognition-exists/ - What building Shippy taught us about building agents
What building Shippy taught us about building agents
- Job hunters are using AI to cheat in interviews, and failing at the office
Job hunters are using AI to cheat in interviews, and failing at the office The Straits Times
- Google Releases LiteRT.js: A JavaScript Binding of LiteRT That Runs .tflite Models in Browsers via WebGPU
Google Releases LiteRT.js: A JavaScript Binding of LiteRT That Runs .tflite Models in Browsers via WebGPU MarkTechPost
- The Unit Economics Of India’s Voice AI Boom
India’s voice AI revolution has revolved around a few key pillars, such as speech recognition accuracy, multilingual capabilities, latency and…
- New York offers Democrats a roadmap for data center fight
New York Gov. Kathy Hochul's first-in-the-nation data center moratorium could provide a playbook for other Democrats confronting one of the midterms' most heated issues. Why it matters: Voters across the country are protesting data centers and looking for politicians to take decisive action to address concerns over the industry's rapid expansion. Until now, no governor had imposed a statewide data center moratorium. Hochul's approach — using executive authority while continuing to negotiate over broader legislation — could become a playbook for other governors. Zoom in: Hochul went the executive order route in order to move as quickly as possible, her staffers say, while continuing to work with the New York legislature. State Sen. Kristen Gonzalez recently pushed a data center moratorium bill through the New York legislature, but instead of signing that, Hochul took executive action. Hochul's staffers said that Gonzalez's legislation raised more complicated policy questions that require further negotiation. For example, Gonzalez said, negotiations will likely include clearer guidelines for public hearings. Gonzalez told Axios that she's "really supportive" of Hochul's approach and worked with the governor's office because she agreed the state needed to act quickly. The fine print: Gonzalez's bill goes further than Hochul's order. It would impact projects above the 20 megawatt level instead of 50, create new electric and water rate classes and require public hearings before future permits are approved. Hochul's staffers say the governor wants to pair her executive order with legislation to ensure hyperscalers don't get tax benefits, while continuing to negotiate with Gonzalez on her bill. Moratorium bills are cropping up across the country. Gonzalez said this will "certainly support" other governors considering proposals that are likely to make it to their desks. Zoom out: The backlash isn't limited to blue states. Texas Gov. Greg Abbott (R) recently called for banning new AI data centers in rural neighborhoods and to require the industry to shoulder more of its infrastructure costs. Between the lines: Hochul is toeing the line between companies and voters, and acting fast without shying away from language viewed by some other Democrats as too politically charged. "Now, no one can accuse New York of fearing innovation," Hochul said at a signing ceremony on Tuesday. "But that said, a lot of Americans — for a lot of them, the specter of unchecked AI brings up fear, anxiety, a lot of worry." Hochul and President Trump have aligned around efforts to have companies pay for the energy they demand. The intrigue: Sen. Bernie Sanders (I-Vt.) and Rep. Alexandria Ocasio-Cortez (D-N.Y.) thrust the idea of a data center moratorium into the national spotlight earlier this year with legislation that would have paused new projects nationwide. Sanders on Tuesday noted the idea, once "radical," is becoming a reality in places like New York. The Sanders-AOC proposal went even further than Gonzalez's bill, and would lift the ban only after Congress passed sweeping AI safety legislation covering issues from civil rights to consumer protection. The bottom line: Whether New York proves to be an outlier or becomes a model, Hochul has given governors a new playbook for responding to the data center backlash.
Score: 45🌐 MovesJul 15, 2026https://www.axios.com/2026/07/15/new-york-democrats-roadmap-data-center-fight - AWS EC2 and AI leader Dave Brown to exit, replaced by Amazon exec and Microsoft vet Dave Treadwell
Dave Brown is leaving Amazon Web Services after nearly 19 years, departing at the end of July for a new role outside the company. Dave Treadwell, a longtime Amazon executive who spent 27 years at Microsoft, will take over AWS Compute and ML Services on Aug. 1. Read More
- OpenAIs smart speaker sounds like a cross between a HomePod and a Furby
OpenAI is reportedly making a portable, screenless smart speaker intended to act as a household AI companion.
- SREs To AI Agents: Prove Yourself Before You Touch Production
SREs To AI Agents: Prove Yourself Before You Touch Production
- Mythos and the Future of AI Cyber Threats
Mythos and the Future of AI Cyber Threats IT Pro
Score: 44🌐 MovesJul 15, 2026https://www.itpro.com/technology/artificial-intelligence/mythos-and-the-future-of-ai-cyber-threats - AI cloud company CoreWeave explores Wall Street playbook to hedge memory-chip price risk
The unusual move underscores how deeply the AI boom has entangled cloud providers with the volatile chip market. To lock in supply amid soaring demand, thanks to a surge in AI infrastructure construction, cloud operators including CoreWeave have signed long-term agreements with memory and storage makers such as Micron and SanDisk.
- AI slop movies are the new direct-to-video cash grabs
This weekend, cinephiles across the world will march to their local theaters to feast their eyes on Christopher Nolan's new adaptation of The Odyssey. It's on track to rake in anywhere between $80-$100 million in just a few days. People are clearly excited to see how Nolan uses cutting-edge filmmaking tech to make the Homeric […]
Score: 43🌐 MovesJul 15, 2026https://www.theverge.com/entertainment/965616/ash-koosha-odysseus-the-fall-foundtain-zero-tilly-norwood - ESET Threat Report: AI boosts cyber attackers’ efficiency
ESET Research has released its H1 2026 Threat Report with statistics from December 2025 through May 2026
- Anthropic official says stopping AI usage is 'the wrong' response to AI cost concerns
Anthropic official says stopping AI usage is 'the wrong' response to AI cost concerns Business Insider
Score: 43🌐 MovesJul 15, 2026https://www.businessinsider.com/anthropic-ai-costs-responses-routers-2026-7 - E2W startup E3 Electric.Ai raises Rs 100 crore in equity, debt
The funding comprises Rs 75 crore in equity and Rs 25 crore in debt, and more than 80% of the capital has already been disbursed, founder and chief executive P Sanjeev told ET. The company plans to use the proceeds for product engineering and commercialisation of its first scooter.
- Cyber, AI and cost control: The priorities defining enterprise IT decisions
Cyber, AI and cost control: The priorities defining enterprise IT decisions Computing UK
- AI schools are multiplying, but the promise of AI tutoring is complicated
Over the past decade, the AI -focused, for-profit Alpha School has grown from one campus in Austin to a growing list of more than 15 schools across the country, including in major cities like New York and San Francisco. Alpha School joins other AI-centric K-12 private schools , like Unbound Academy and the Khan Lab School , that parents can now choose for their children, paying anywhere from $40,000 to $75,000 per year in some cases. Some advocates of AI-driven education use slogans like “ School is broken, and we’re here to fix it .” MacKenzie Price, one of Alpha School’s cofounders, has described her frustration with the “ one-size-fits-all ” model of schooling, in which all students study the same material but often learn at different speeds. Sal Khan, founder of the online educational platform Khan Academy, has also argued that AI could give every student an individual tutor that’s responsive to their specific needs . And Bill Gates is among the tech thought leaders who have speculated that AI will replace many teachers within the next decade. The argument for more personalized education may be a good one in some cases. But it also risks creating a false dichotomy where all AI programs are seen as responsive and motivating, and all classroom teaching is rote lecturing. As a scholar of education policy who does research on AI and teachers, I recognize that reality is far more complicated. The promise of AI tutoring Alpha School replaces traditional, in-person instruction for K-12 students with personalized AI tutoring that condenses reading, math, and other subjects into a two-hour study period. The two-hour AI tutoring block is supplemented with in-person workshops and what the school calls “coaches” or “guides,” who are not necessarily licensed teachers. These sessions can focus on nonacademic life skills, like public speaking and entrepreneurship, as well as art and physical education. When it comes to the AI portion of this education, it’s clear that there are benefits to personalized tutoring. A major review published by the National Bureau of Economic Research in 2020, for example, showed that many different kinds of tutoring provided by humans produced consistent learning gains across subjects and age ranges. A 2026 study by the Brookings Institution showed that AI improves computer-assisted tutoring by allowing students to interact in everyday language with the program. Generative AI can also modify the tutoring session, based on a student’s progress and challenges with the material. So far, though, there is no clear evidence that AI or other kinds of computer-based tutors are superior to human tutors , though in many cases they might be cheaper. Instead of trying to replace teachers with AI, I think that a more promising strategy is to support teachers in using AI to become better educators. Good tutoring supports classroom learning One-to-one tutoring with a human teacher, parent, or paraprofessional has been shown to be very effective in raising student test scores and to improve student learning overall. “ High-impact ” tutoring that involves regular, long-term sessions in small groups held within schools has been shown to be particularly effective. As Isabelle Hau , executive director of the Stanford Accelerator for Learning, writes, young students need to develop strong social skills, what she calls “ relational intelligence ,” to flourish in social settings. To be effective, human tutoring needs to meet certain criteria. For example, tutors and their students should meet consistently, and the tutors should be skilled and make sure that what they are teaching is fully integrated with the student’s curriculum . Relationship-building between tutors and students is also critical for getting students to stay more motivated and focused on their work at hand. Can AI tutors replace human ones? Computer-assisted instruction and intelligent tutoring systems have been around since the 1970s . These systems are designed to provide instruction that adapts to learners as they progress through a series of exercises, problems, or questions. They have been shown to increase student achievement when used as a supplement to classroom learning . However, extensive analysis of different studies found that while these systems increased student achievement, when compared with classroom instruction or students who rely on textbooks or workbooks to study, there was no significant difference in learning between computer tutors and human tutors . More recently, a 2025 study found significant positive effects of AI tutoring on student achievement across subjects and different grade levels. In another widely cited study on AI tutors, researchers in 2025 compared the effect of active learning in class with a specially designed AI tutor they had built for an introductory class in physics at Harvard . Students who used the AI tutor said they learned the material faster and felt more motivated to learn, compared with students who received specialized in-class instruction. In this study, not only were the students already highly motivated and had good study skills, but the AI tutor was built by the same professors who taught the course. AI-supported instruction Claims that AI tutoring is superior to classroom instruction are only looking at part of the evidence . But AI tutors could indeed be useful in some cases, if they are used to support human teachers in their work. One 2024 study showed student improvement in middle schools in low-income areas when human tutors had access to AI tutoring support. A 2026 study showed that when teachers use AI for lesson plans, experienced teachers tend to critically revise content to better link it to the overall curriculum. Instead of considering a future where AI tutors replace human teachers, I believe that we need to think about what kind of support and professional development can help teachers make the most of these new and complicated tools. Gerald K. LeTendre is a professor of educational administration at Penn State . This article is republished from The Conversation under a Creative Commons license. Read the original article .
- What 80% AI-written test pipelines actually cost
What 80% AI-written test pipelines actually cost InfoWorld
Score: 42🌐 MovesJul 15, 2026https://www.infoworld.com/article/4196873/what-80-ai-written-test-pipelines-actually-cost.html - Agent identity: Agents that act and appear as themselves
Enterprise agents gain scoped credentials, persistent presence, and audit trails to act securely across Slack, Jira, GitHub, and more.
- The Hidden Cost of Letting AI Agents Write Your Tests
I asked Claude to write tests for a payment processing module last week. It generated 47 tests in under two minutes. Every single one… Continue reading on Towards AI »
- Daimler Truck Finance: Navigating the Freight Downturn & AI Boom
SummaryView Transcript After four years of shrinking capacity, is the trucking industry finally seeing the end of the longest freight recession on record? Daimler Truck Financial leaders Tobias Waldeck and Kevin Bangston share their insights on market recovery, fleet investment trends, and how new EPA regulations and the AI data center boom are reshaping the […] The post Daimler Truck Finance: Navigating the Freight Downturn & AI Boom appeared first on FreightWaves .
Score: 42🌐 MovesJul 15, 2026https://www.freightwaves.com/news/daimler-truck-finance-navigating-the-freight-downturn-ai-boom - The Irish graduate jobs most affected by AI
Recruitment experts say while there is a ‘tightening’ of graduate hiring, AI is also leading to a more diverse range of entry-level roles
Score: 41🌐 MovesJul 15, 2026https://www.irishtimes.com/business/work/2026/07/15/what-effect-is-ai-having-on-entry-level-roles-in-ireland/ - Strategic Autonomy in the Age of Artificial Intelligence
This essay argues that the rapid advance of AI will frustrate ambitions for strategic autonomy.
- Mediagenix Introduces Trusted Agentic AI Operating Model for the Real-Time Media Enterprise
Mediagenix Introduces Trusted Agentic AI Operating Model for the Real-Time Media Enterprise USA Today
- AI remains unreliable as one-size-fits-all code security tool, comparison finds
For a paper published in the International Journal of Applied Cryptography, a team compared 11 leading large language models (LLMs) for software security. They found that no single system consistently outperforms its rivals in detecting vulnerabilities. This, they suggest, means organizations must select such tools according to the specific software they are analyzing.
- Conversation Intelligence: The complete guide for 2026
Comprehensive guide covering conversation intelligence trends and tools for 2026.
- Microchip Advances Neural Network Implementation with VectorBlox 3.0 Accelerator SDK
Deploying AI inference in power‑constrained and mission‑critical environments such as aerospace and defense systems requires solutions that balance performance, efficiency, reliability and ease of development. To better manage these challenges, Microchip Technology (Nasdaq: MCHP) has released the VectorBlox™ 3.0 Accelerator Software Development Kit (SDK) to help simplify FPGA‑based AI implementation and speed time‑to‑market. Offered to developers free of charge, VectorBlox […] The post Microchip Advances Neural Network Implementation with VectorBlox 3.0 Accelerator SDK appeared first on CXOToday.com .
- Context is becoming AI’s most misunderstood word
If you spend enough time in Silicon Valley AI circles, you’ll hear the same message over and over again: AI needs context. The statement is broadly true. The problem is that “context” has become one of the least precise terms in the industry. Depending on who is using it, context can mean documents, dashboards, reports, metadata, business rules, policies, transaction histories, CRM records, knowledge bases or institutional expertise. The word has become a catch-all for virtually any information that might be made available to a model. As a result, many organizations have started treating context as a volume problem. Conversations quickly turn to larger context windows, additional data sources and broader system access, while far less attention goes toward determining whether that information actually improves the quality of the outcome. What we’re seeing in practice suggests a different way of thinking about the problem. The organizations making the most progress with enterprise AI are not necessarily the ones exposing the largest amount of information to their systems. They are the ones spending the most time understanding which information should influence a decision, which information should not and how to ensure that business logic is applied consistently. That distinction matters because the industry is beginning to repeat a mistake enterprises already made once before. Context has become the new ‘big data’ For much of the last two decades, organizations operated under the assumption that collecting more data would naturally produce better decisions. Massive investments were made in data warehouses, reporting platforms, analytics systems and business intelligence tools. Those investments created tremendous value, but they also exposed an important reality: Collecting information and creating clarity are not the same thing. Today, AI is heading down a similar path. Many enterprise AI projects measure progress by counting how much information a model can access. More documents become better than fewer documents. More systems become better than fewer systems. Larger context windows become better than smaller ones. The conversation often assumes that quantity and quality move together. Well, they don’t. According to Salesforce research , only 35% of business leaders say they are completely satisfied with their organization’s ability to use data effectively despite years of investment in data infrastructure and analytics. Enterprises learned long ago that information alone does not create understanding. The same lesson applies to AI. When a model gains access to five versions of the same metric, conflicting definitions of a business process or documentation that has not been updated in years, it does not magically resolve those inconsistencies. It consumes them. More context can just as easily increase ambiguity as reduce it. Simply exposing more information to a model does not guarantee better outcomes. What matters is whether the information available to the system helps it make the right decision at the right time. Most AI failures are actually context failures One of the more interesting things we’ve observed over the past year is how many AI projects are blamed for problems that have very little to do with AI. The model answers a question incorrectly, and the immediate assumption is that the model failed. In reality, the underlying issue often sits elsewhere. The organization may have multiple definitions of the metric being requested. Customer information may exist across several systems with conflicting values. Business rules may be documented in one location, partially implemented in another and understood differently by different teams. In many deployments, the issue is not that the AI lacks information. The issue is that it has access to several competing versions of the truth. Anyone who has worked inside a large enterprise will recognize the pattern. Revenue means one thing to finance and something slightly different to sales. Product usage metrics evolve over time. Operational processes change while documentation remains frozen. Human employees learn how to navigate these inconsistencies through experience and institutional knowledge. AI systems inherit them immediately. This is why the conversation around context often misses the point. The challenge is not simply providing more information. The challenge is determining which information should be trusted, how conflicts should be resolved and what business logic should govern the final answer. A single trusted source can be more valuable than a hundred loosely connected ones. A clearly defined rule can be more useful than thousands of pages of documentation. The quality of the context matters far more than the volume. Access does not create trust Many organizations can tell you exactly how their AI systems retrieve information. They can explain retrieval pipelines, vector databases, ranking systems, semantic search architectures and context windows in extraordinary detail. Far fewer can explain how they determine whether the answers produced are consistently correct. That gap becomes especially important in enterprise environments where the cost of an incorrect answer can be substantial. A sales leader making a forecast, a finance team evaluating performance or an operations executive making a resource allocation decision does not care how many documents were retrieved. They care whether the answer is right. Trust has always been one of the hardest problems in enterprise data. According to Accenture research on data trust and decision making , only about a quarter of employees report high confidence in their organization’s data when making decisions. That challenge does not disappear when AI enters the picture. If anything, it becomes more visible. Organizations frequently measure access because access is easy to quantify. Reliability is harder. Reliability requires understanding whether an answer remains consistent across users, across prompts, across time periods and across changing business conditions. It requires understanding whether the same question produces the same answer and whether that answer reflects the business logic the organization intends to enforce. Those are fundamentally different measurements, and they point to a different definition of success. Context requires measurement One reason this problem is becoming more pronounced is that enterprises accumulate information far faster than they eliminate it. New systems are added, new reports are created, processes evolve. Teams develop local definitions and specialized workflows. Documentation grows continuously, while very little of it gets removed. Over time, organizations build large collections of information that contain years of historical decisions, exceptions, workarounds and competing interpretations. We’ve yet to encounter an enterprise that doesn’t have some version of this problem. That reality turns context into an operational challenge rather than a technical one. Simply connecting AI systems to enterprise information does not improve the quality of that information. In some cases, it exposes longstanding inconsistencies that were previously hidden by human interpretation and tribal knowledge. Gartner has long identified poor data quality as one of the most significant obstacles to successful analytics and AI initiatives because bad inputs inevitably produce unreliable outputs, regardless of how sophisticated the technology becomes. As AI becomes more deeply integrated into business operations, organizations will need new ways to evaluate the context their systems rely on. They will need visibility into how information is being used, where definitions conflict, which sources are trusted and how context quality affects outcomes. Context cannot be treated as a static asset. It must be measured, monitored and improved over time, just as organizations measure the quality of the models and applications built on top of it. The shift from access to reliability The industry has spent the last several years focused on access. How do we connect models to enterprise systems? How do we expose organizational knowledge? How do we give AI visibility into the information people use every day? Those questions were important because they represented genuine technical barriers. Today, many of those barriers are disappearing. Most enterprises can already connect AI systems to data warehouses, applications, dashboards, documents and knowledge repositories. The conversation is beginning to shift toward a more difficult problem: Determining whether those connections actually produce outcomes people trust. That is where the next phase of enterprise AI will be decided. Organizations that treat context as a quantity problem will continue adding more information and hoping accuracy improves. Organizations that treat context as a quality problem will focus on trust, consistency, governance and outcome reliability. The difference between those approaches may sound subtle, but it has enormous implications. One produces systems that can access information. The other produces systems that people are willing to use to make decisions. And in the enterprise, that distinction is ultimately what matters. This article is published as part of the Foundry Expert Contributor Network. Want to join?
Score: 40🌐 MovesJul 15, 2026https://www.cio.com/article/4196864/context-is-becoming-ais-most-misunderstood-word.html - Could now finally be a good time to buy an AI PC? This report says so
Analysts say AI PCs could run small models and local AI workflows without needing to rely on expensive, unpredictable cloud compute.
Score: 40🌐 MovesJul 15, 2026https://www.techradar.com/pro/could-now-finally-be-a-good-time-to-buy-an-ai-pc-this-report-says-so - This software team will charge you $10,000 a week to remove all AI-generated code from your systems — and use AI to do it
For $10,000 a week, a three-man team will use AI coding agents to find lengthy AI generated codebases within your system’s internal applications, and trim fat.
- Home Office proposes AI tooling to accelerate digital disclosure
The Home Office has accepted key recommendations from an independent review of how digital material submitted to courts can be managed more effectively. London barrister Jonathan Fisher, who specialises in corporate and financial crime, produced the Independent review of disclosure and fraud offences , which was presented to the UK Parliament in March 2025. His review covers a number of areas where the disclosure process for digital material across the criminal justice system has caused bottlenecks. At the time, Fisher noted that the largest investigation case on the Serious Fraud Office (SFO) system has 48 million documents (6.5 Tbytes of data). “With this volume of digital material, it is inconceivable that the totality of unused material generated in the investigation can be accurately reviewed and scheduled by investigating officers manually, in the traditional way,” he said. “It is also a gross waste of resources for investigating officers to spend time on banal and unproductive activity.” Fisher recommended an assessment of safe and ethical use of advanced technology in an open and transparent manner. “As has been the case with many previous technological advancements, the use of technology has the potential to reduce administrative burden and increase accuracy,” he said. “We must have a system that takes a pragmatic approach to disclosure, using ethical, secure and accurate advanced technology to streamline the processing, redaction and scheduling of large volumes of digital material. This will effectively free up resources to focus on the complex aspects of investigation and prosecution rather than arduous administration, increasing the speed of justice for both victims and defendants.” Under the current system, police officers manually process and provide a written summary for every file that could be relevant to an investigation. The new regime proposed by the Home Office will mean that police officers will be able to use technology to identify, sort and compile millions of files that are currently reviewed manually. Graham McNulty, director of the Serious Fraud Office, said: “Modern fraud, bribery and corruption cases involve vast amounts of digital data, and our disclosure regime must keep pace with that reality. I welcome the government’s response to disclosure reform, which marks a positive step forward. “I’m particularly pleased that the government will be exploring opportunities to pilot a new intensive disclosure regime, and look forward to continuing to work with others to bring disclosure practice fully into the digital era.” According to an estimate from the Policing productivity review , officers spent approximately 532,000 hours in 2022/23 undertaking disclosure work and building case files which were later assessed by the Crown Prosecution Service (CPS) as requiring no further action. PoliceAI, the government-backed initiative launched in June, aims to free up an estimated six-million hours of police time per year by 2028 – equivalent to 3,000 extra officers. Through PoliceAI, the Home Office said it will fund piloting of AI tools that are capable of automatically generating summaries of digital material, which it stated would save investigators countless hours spent carrying out administrative tasks, with a view to scaling across all police forces in 2027. Al Murray, interim director of PoliceAI, said: “The disclosure process is an essential safeguard in our justice system, but the scale of digital evidence involved in modern investigations means we need modern solutions. Used responsibly, AI can help officers and investigators manage vast amounts of material more efficiently, allowing them to spend more time supporting victims, pursuing offenders and exercising the professional judgement that technology can never replace.” Minister for policing and crime prevention Sarah Jones said: “Police officers are wasting thousands of hours trawling through phones, emails, messages, videos and cloud storage because of outdated regulations. By embracing AI and new technology responsibly, we will boost productivity, bring policing into the 21st century and free officers to focus on the frontline.” Read more UK policing stories UK police to upgrade illicit asset recovery system : A new system is being developed by the Police Digital Service and NEC Software Solutions to help manage the recovery of criminal assets, replacing 20-year-old legacy infrastructure. Police facial recognition must be strictly limited: Digital rights groups map out ‘minimum, necessary’ human rights protections to be included in UK government’s upcoming legal framework for police facial recognition.
Score: 40🌐 MovesJul 15, 2026https://www.computerweekly.com/news/366645936/Home-Office-proposes-AI-tooling-to-accelerate-digital-disclosure - Model Routing Is Simple. Until It Isn’t.
Model Routing Is Simple. Until It Isn’t.
Score: 40🌐 MovesJul 15, 2026https://huggingface.co/blog/ibm-research/model-routing-is-simple-until-it-isnt - 😼 Google wants an AI referee
DeepMind pushed model oversight as AI's power bill grew.
- Don’t Let Claude Grade Its Own Homework
Cross-provider PR review with Codex in GitHub Actions, and why a second opinion from a different lab beats any self-review The post Don’t Let Claude Grade Its Own Homework appeared first on Towards Data Science .
- WAIC 2026 Preview: Domestic AI Supernodes Take Center Stage as AI Computing Enters the System Era
WAIC 2026 showcases China domestic AI supernode computing clusters with Sugon Dawn, Hygon processors, and 100K-card-scale systems, marking a shift from chip competition to system-level computing.
- Cursor Is Reinventing Itself as SpaceX Deal Looms
Cursor Is Reinventing Itself as SpaceX Deal Looms The Information
Score: 40🌐 MovesJul 15, 2026https://www.theinformation.com/articles/cursor-reinventing-spacex-deal-looms - This Robot Was Designed to Look Like Furniture. That Could Be Its Killer Advantage
The two Apple veterans behind Weave are hoping their $7,999 robot is helpful in homes, even if it can only do two things.
- Keeping AI agents running in the business chat
As SMBs run more of their operations through WhatsApp and Telegram, Hostinger’s infrastructure helps keep AI agents accessible, connected, and always
- 1Password launches AI token spend management
Amid soaring costs, the Toronto firm aims to help clients get a handle on their AI usage. The post 1Password launches AI token spend management first appeared on BetaKit .