AI News Archive: May 29, 2026 — Part 6
Sourced from 500+ daily AI sources, scored by relevance.
- Paystack launches AI-powered dashboard to help African businesses understand payments
More than 300,000 organisations — including startups, enterprises, SMEs
- New Data Formulator 0.7: AI analytics make enterprise data easier to explore
The post New Data Formulator 0.7: AI analytics make enterprise data easier to explore appeared first on Source .
Score: 37🌐 MovesMay 29, 2026https://www.microsoft.com/en-us/research/blog/data-formulator-0-7-ai-powered-data-analytics-for-enterprise-data/ - AI recruitment: How your next job interview could be with a bot
Almost 60% of German respondents to a recent survey said they have conducted an interview with AI. Here are some tips for how to ace it.
Score: 37🌐 MovesMay 29, 2026http://www.euronews.com/next/2026/05/29/ai-recruitment-how-your-next-job-interview-could-be-with-a-bot - Are you afraid AI will take your job? Gov. Gavin Newsom wants to hear from you
Are you afraid AI will take your job? Gov. Gavin Newsom wants to hear from you The Mercury News
- Snyk creates operational roadmap for the AI governance maturity model
The way organizations manage AI is changing. The shift is from simple models to agents that act on their own — sometimes for better, sometimes for worse. To handle this, good AI governance means an organization must continuously know, control, and prove what its AI systems are doing once they are running. Many groups still … continue reading The post Snyk creates operational roadmap for the AI governance maturity model appeared first on SD Times .
Score: 37🌐 MovesMay 29, 2026https://sdtimes.com/ai/snyk-create-operational-roadmap-for-the-ai-governance-maturity-model/ - Enterprise AI is entering its accountability era
Enterprise AI is entering its accountability era Techcircle
Score: 36🌐 MovesMay 29, 2026https://www.techcircle.in/2026/05/29/enterprise-ai-is-entering-its-accountability-era - Are consumers ready for humanoid robots?
Should we be worried about the future?
- The AI agent bottleneck isn't model performance — it's permissions
Enterprise AI agents are stalling — not because of model performance, but because of permissioning. Every agentic workflow eventually hits the same wall: what is this agent allowed to touch, on whose behalf, and how does the system know? Workday's answer is to make its existing system of record the governance layer for agents. Gerrit Kazmaier, the company's president for product and technology, told VentureBeat in an interview that customers often struggle when they cobble together solutions for their agents. “Sana makes sure the integrity of the approvals and security model is always adhered to,” Kazmaier said. “Frankly, that’s where we see customers struggling when they try to build do-it–yourself AI by just accessing raw data, so the richness of the security model gets lost, and the results become overly broad.” Workday, which launched Sana in March , expanded its partnership with Google to bring its Sana agent system of record to the Gemini Enterprise — so agents built on Sana are also discoverable there. Architecting accuracy Kazmaier said the biggest hurdle they faced was ensuring agent accuracy, especially for HR and finance users. “Almost right is not acceptable,” Kazmaier said. “Think about paying people correctly, closing the books or managing work schedules reliably.” Accuracy is harder to evaluate here than in most AI contexts. Policy configurations, role-based security, and organizational hierarchies are deeply interrelated — a small error compounds. And unlike most generative AI outputs, HR and finance queries often lack a correction loop. By the time a paycheck processes incorrectly or an interview is scheduled wrong, the damage is done. Workday addressed this by building Gemini in as its base reasoning layer, then adding its context engine and business process logic on top. Workday also added verification and classification models that “interrogate” outputs before execution. Accuracy and identity, it turns out, are the same question: does the system know enough about the agent, the authorizing human, and the current state of the record to act correctly? Workday’s advantage is that it can infer its customers' organizational structures from the data they provide. Already, third-party identity providers like Okta verify their information by checking Workday, so its context is the system of record for many enterprises. Kazmaier said the Sana Self-Service Agent uses Gemini as the conversational surface to trigger the workflow. The user is then authenticated and authorized through Workday’s identity and security model. Sana agents will only act on behalf of that user and work within their current permissions. Audit trails follow the same logic: Gemini retains only interaction logs, while the main audit remains within Workday and its customer. For many practitioners in the HR and finance space, the permission and governance layer in the agent system of record is key in regulated spaces. “It has to live in the system of record, that’s not a preference, that’s the only way it works,” said Dan Obendorfer, director of product at Würk, in an email to VentureBeat. “If your permissions are defined somewhere outside of where the data actually lives, you’ve already lost.” Kadan Stadelmann, chief technology officer and co-founder of Compance.AI, made the same point separately. “Without agent ownership, performance, costs or actions, chaos ensues.”
Score: 36🌐 MovesMay 29, 2026https://venturebeat.com/orchestration/the-ai-agent-bottleneck-isnt-model-performance-its-permissions - Rivian is pretty sure customers want AI, not Android Auto
'We see cars becoming more and more AI-defined.'
Score: 36🌐 MovesMay 29, 2026https://www.androidauthority.com/no-android-auto-carplay-for-rivian-3672638/ - Why Your AI Agent is a Black Box and How to fix it With OpenTelemetry
Why Your AI Agent is a Black Box and How to fix it With OpenTelemetry DevOps.com
Score: 36🌐 MovesMay 29, 2026https://devops.com/why-your-ai-agent-is-a-black-box-and-how-to-fix-it-with-opentelemetry/ - How AI Is Removing the Bidding Capacity Ceiling in Construction
How AI Is Removing the Bidding Capacity Ceiling in Construction india.entrepreneur.com
Score: 36🌐 MovesMay 29, 2026https://india.entrepreneur.com/business-news/how-ai-is-removing-the-bidding-capacity-ceiling-in-construction - AI Researchers, Ask Yourself These 6 Questions to Strengthen Your Moral Muscles
By Max Tegmark & Meia Chita-Tegmark Of course you have moral principles – but how often do you use them? I, Meia, am a professor doing psychology research, and I can tell you that most bad outcomes are caused not by lack of moral principles, but by them not being activated. I, Max, am a professor doing AI research, and I can tell you that your choices as an AI researcher truly matter, because you’re helping build what will become the most powerful technology ever: AI will gain the potential to bring either unprecedented health, prosperity, liberty, dignity and empowerment, or a race to replace our jobs, our relationships, our decision-making, our power and even our species. Hardly a day goes by without the AI community facing moral decisions, on topics ranging from AI companions to surveillance, hacking and military use. Many top AI companies are fighting lawsuits about everything from data centers to AI safety, and Anthropic is in a prolonged showdown with the Pentagon. So for all you AI researchers out there, here’s a handy checklist to tone up your moral strength. 1. Do you have red lines? Is there any action that you find so morally unacceptable that, if the organization you work for takes it, you’ll quit? Or take some other pre-determined costly action, say whistleblowing? Such actions are your moral red lines. For example, Rosa Parks got fined and fired for her civil disobedience against segregation, Vasily Arkhipov was criticized after vetoing a Soviet nuclear strike against the US, and Edward Snowden ended up in exile for mass surveillance whistleblowing. Many AI researchers have left top AI companies that crossed their red lines, including Daniel Kokotajlo, who risked almost $2M in equity by quitting OpenAI without signing a non-disparagement agreement. What are your red lines? 2. Have you written them down and shared them? Both George Washington and Benjamin Franklin wrote down moral guidelines for themselves, with Franklin grading his own performance weekly. This is a powerful tool for avoiding the boiling frog effect, protecting your red lines against gradual erosion as in the examples at the end of the next section. Sharing them with loved ones or online adds social pressure to stick to them. For each red line, make sure to write down what action you commit to taking if it is crossed. You can click here to list your red lines (we will only share them with your permission). 3. Have you resisted moral disengagement? To further strengthen your moral muscles and ensure that your red lines don’t move, it’s helpful to know what failure mechanisms to watch out for. Disengaging your muscles makes you weak–and this applies to your moral muscles as well. So let’s look at moral disengagement mechanisms identified by Albert Bandura, one of the most impactful psychologists of all time. This will help you spot them and fight them when your red lines get pressured by your company, your social circle, the temptation of personal gain, or the desire to feel good about yourself. Displacement and diffusion of responsibility: You’ll feel better if you or others convince you that you’re not really responsible for the harm: the real decision maker is leadership, investors, the market, geopolitics, or history (“this technology is inevitable”). When AI work is distributed across large teams, everyone feels less accountable for the collective outcome. “I’m just a researcher” or “I was just doing my job” are archetypical excuses identified by the influential political theorist Hannah Arendt. The satirical musician Tom Lehrer sums it up in this hilarious song about the rocket scientist who switched allegiance from Nazi Germany to the US: “Once the rockets go up, who cares where they come down – that’s not my department, says Wernher von Braun” . For example, an Anthropic researcher reading about how their Claude AI may have been implicated in killing over 150 Iranian school girls, in one of the worst US-caused civilian bloodbaths since the Vietnam War, may be tempted to tell themselves that they’re blameless because only management is responsible for selling their tools for military targeting. Word games: Both Bandura and Arendt highlight how subtle word choices can reframe what’s moral. We are all familiar with military euphemisms such as “servicing a target” for bombing, “collateral damage” for civilian casualties and “enhanced interrogation techniques” for torture, but AI jargon is full of analogous word games, often encouraged by financially interested parties. The most basic game is “euphemistic labeling”: replace morally vivid language with positive or emotionally flattened terminology. Researchers are not “helping build systems that may displace workers, manipulate users, centralize power, or heighten existential risk”; they are doing “capabilities research”, “model improvement”or “benchmark progress”. Training on copyrighted data becomes “freedom to learn”. Unpopular data centers become “AI infrastructure”. Firing or deskilling workers becomes “productivity gains” and “Lobby against accountability” becomes “reduce friction”. Please practice using neutral words like “company” instead of “lab” (which sounds cool and innocent) and “AI system” instead of “AI model” (which sounds harmless). Bandura’s point is that euphemism does not merely soften tone; it weakens conscience. Another word game is blame attribution, where critics become the problem, say “doomers,” “Luddites”, “opportunistic politicians”, “ignorant journalists” or “anti-tech Europeans”. Once opponents are blamed for irrationality or bad faith, the AI researcher feels less obligated to treat criticism as morally serious. A third word game is soft dehumanization: the unemployed programmer, the individual copyright infringement victim and the chatbot suicide child disappear into categories such as “the labor market”, “creatives” and “edge cases”. The more harms are discussed statistically rather than personally, the less moral pain is triggered. Selective moral self-exemption: It’s tempting to keep strong moral standards in general, but carve out an exception around the domain from which you benefit most: an AI researcher may be passionately ethical about injustice in the abstract, while suspending those same standards when judging their own employer, AI, salary or stock grant. Advantageous comparison: It’s tempting to compare yourself only to worse actors: “At least I’m not at the most reckless lab.” “At least I’m not working on autonomous weapons.” “At least I care about alignment.” That lets you feel ethical without asking whether your own conduct is acceptable in absolute terms. Moral justification: For those acknowledging that they’re causing current harm, it’s tempting to justify it as serving a noble mission, say “helping democracy prevail”, “creating universal abundance” or “making sure that safety has a seat at the table” – without seriously questioning whether those lofty goals are credible, or whether there’s another way to accomplish them with less current harm. These moral disengagement techniques can be very powerful when combined and escalated: Enron executives gradually escalated from minor financial manipulations, justified as necessary for company survival and diffused through leadership directives, to massive fraud like hiding debt. Bernie Madoff started with small return fudges rationalized as client aid, then displaced blame onto markets and dehumanized victims, leading to a $65 billion fraud through incremental moral disengagement. In the Vietnam War, soldiers obediently followed orders in a “just war”, starting with minor transgressions that escalated to massacres like My Lai through diffused responsibility and victim dehumanization. The frontier AI researcher’s signature Bandurian mantra is “I’m not a well-paid participant in a harmful race; I am a responsible, realistic, morally serious person helping guide inevitable progress” . But is the race to replace truly inevitable, given polling finding it wildly unpopular, or is it a Bandurian excuse and self-fulfilling prophecy? 4. Do you maintain situational awareness? Do you actively research whether your red lines are being crossed? This includes investigating the indirect consequences of what your organization does. Hannah Arendt wrote about “the banality of evil”. arguing that the greatest harms are often done not by malice, but by obedient and conscientious technocrats who don’t think about the bigger picture. We talked above about taking known harms and using word games to downplay and reframe them as manageable, transitional or outweighed by upside. But there’s also another powerful moral disengagement technique: staying conveniently ignorant by not putting in the effort to know about the harms you’re contributing to in the first place. Ignorance is a bad excuse if you could have found out by looking into it: German Chemist Bruno Tesch was convicted and executed in 1946 for supplying Zyklon B gas to Auschwitz-Birkenau despite claiming he didn’t know what it would be used for. So please ask obvious questions regularly. For example, which if any red lines does your organization have? Is it actively lobbying against AI safety legislation that you support? Have you looked it up in the AI safety index? How are its products used? If you work for Google or OpenAI, have you skimmed any of the lawsuits against your company for alleged chatbot-linked suicide? If you work for Anthropic, how much do you know about that girl school strike ? Ironically, thanks to modern LLMs, there’s really no excuse for not knowing about things like these, since they’re just a prompt away. For example, you can try this monthly: "Please make a list of morally questionable/controversial behavior by [MY COMPANY] in recent years, including a) controversial use of its tools (say for suicide, crime, surveillance or weapons), b) harm allegedly caused by its tools, c) alleged lies or broken promises by the company or its leadership, d) perverse incentives for the company to pursue profit over what truly benefits humanity." These are the ChatGPT responses we got for Anthropic, Google, OpenAI, Meta and xAI on March 29 2026. 5. Do you make noise internally? If you learn about something that’s close to one of your red lines, then ask questions internally to find out more. Although there were historical situations where criticizing one’s organization could get one killed, doing so in an AI company today is unlikely to even get you fired – and why would you want to keep working for a company that can’t handle respectful questions about your red lines? Most even have whistleblowing policies that protect you (see page 99 here in the AI Safety Index). If what you find out is unacceptable but you’re not ready to quit, then make noise internally: explain why to colleagues and superiors, and push hard for change. Don’t be like one of the engineers who realized that the cold weather could cause catastrophic O-ring failure in the Challenger Space Shuttle, and later regretted not speaking up forcefully. If you’re in the safety team and don’t know people in the lobbying team or those who make launch decisions: make a sincere effort to connect with them and educate them – don’t become a poster child for bystander syndrome. 6. Do you make noise externally? Taking a public stance that challenges your own organization can help in many ways , from helping it improve voluntarily to catalyzing external forces that pressure it (and its competitors) to improve. This doesn’t mean you need to risk exile like Edward Snowden: there are many recent cases where AI researchers have gotten away with well-argued criticism of their company without any retaliation whatsoever. What consequences would you face if you publicly criticized your organization or revealed harmful or illegal behavior? Most US AI companies have a whistleblower policy (see above); please read yours! In addition, a simple search (just don’t do it with your own company’s LLM… :-) will show you many reputable whistleblower organizations offering help with everything from legal support to financial aid should you get fired or sued. So having read this, how would you rate your moral muscles? How many moral disengagement techniques did you recognize in yourself, and how strong has your research been on potential harms caused by your company? Please don’t feel disheartened if you scored low despite meaning well. Instead, think of it as going to the gym for the first time, and discovering that you can’t even bench 50 pounds: muscles need to be used to get strong, and this 6-step plan can strengthen your moral muscles in no-time – and you’ll start feeling really great looking at yourself in the mirror! Discuss
Score: 35🌐 MovesMay 29, 2026https://www.lesswrong.com/posts/w4ynu8L57dJP29p5c/ai-researchers-ask-yourself-these-6-questions-to-strengthen - Most teachers use AI but often receive no formal guidance, poll shows
Most teachers use AI but often receive no formal guidance, poll shows The Washington Post
- It’s Time to Turbocharge Your Flywheel. These Are the Best AI Marketing Tools, According to Experts
The right tech stack can help you get to new customers faster and more efficiently.
- Lawyers, meet your AI 'twin'
Lawyers, meet your AI 'twin' Reuters
Score: 35🌐 MovesMay 29, 2026https://www.reuters.com/legal/litigation/lawyers-meet-your-ai-twin-2026-05-29/ - AI-assisted journalism needs disclosure. Here’s mine
Hello again, and welcome back to Fast Company ’s Plugged In . Journalism might not be the world’s single most important job, but it does have one unique distinction: 100% of journalists are interested in it. So it’s no surprise that the profession is the subject of an outsize percentage of articles about the myriad ways AI is changing our world. I do admit, however, to be taken aback by how much of that coverage involves tales of journalists being embarrassed by entirely avoidable AI-fueled gaffes. On May 19, for example, Benjamin Mullin of The New York Times reported that The Future of Truth: How AI Reshapes Reality , a new book by Steven Rosenbaum, executive director of the Sustainable Media Center (and a Fast Company contributor ), contained at least five quotes that appeared to have been manufactured, mangled, or misattributed by AI. Rosenbaum, who told Mullin he took “full responsibility” for the errors, is hardly the only writer to let hallucinated sound bites slip into their work. Earlier in May, The Times itself issued a correction for an article that had attributed an imaginary quote to a Canadian politician. Still, given the subject of Rosenbaum’s book, its AI fabrications couldn’t have been more freighted with irony. Each time I read about some journalistic AI mishap, I ask myself whether I’m at risk of being similarly ensnared. In this case, I’m confident the answer is no, for the simple reason that I never insert anything from a chatbot’s response directly into a story draft. Had a chatbot offered up a punchy quote from reporter/podcaster Kara Swisher —as one apparently did for Rosenbaum—I wouldn’t have assumed it was real unless I could trace it back to its source. That’s what I might quote. If every journalist resisted the temptation to cut and paste algorithmically generated text, far fewer of them would have AI blow up in their faces, Wile E. Coyote-style . But I don’t expect the self-owns to dwindle anytime soon. Indeed, they may proliferate as more writers succumb to AI’s siren call. Writers caught in old-school literary transgressions such as plagiarism have often deflected blame onto their tools (such as cryptic handwritten notes ) or other people (say, overburdened research assistants ). Blaming a seemingly superhuman technological factotum may be even more tempting. Though Rosenbaum admitted culpability for his book’s errors, he also told The Atlantic ’s Will Oremus that he felt “seduced and betrayed” by AI. As well as painfully underlining what can happen when someone gets overconfident about the competence of LLMs, the spate of faulty journalism has convinced me that transparency is in order. Media creators and consumers alike might be better off if AI’s role in reporting and writing were more clearly spelled out. With that in mind, here’s how I’m using the technology—and, crucially, not using it. For better or worse, my wordsmithing is my own, except for the welcome quality control it receives from my Fast Company editorial colleagues. To me, thinking about something and writing about it are pretty much the same act; I wouldn’t know how to turn part of the job over to an algorithm. Even using the technology to brainstorm headlines—which I’ve done on rare occasion —feels like more work than it’s worth. I do cheerfully admit to calling on AI in a variety of ways as I turn raw research into polished prose. In the past, I used Rev to transcribe interviews, Google’s NotebookLM to summarize them, and Grammarly for proofreading assistance. More recently, I’ve replaced them with similar features I built into the bespoke word processor I vibe-coded just for my own use. They’re valuable, but any impact on the finished product is ultimately modest. I also use chatbots, especially Gemini, every day. My goal is not to get answers from a bot (which I wouldn’t fold into my work under any circumstances) but to point my way toward original, human-written sources, often on arcane topics. The bot is the beginning of the journey, not its destination. So far, I’ve found that AI’s greatest gift to my reporting and writing hasn’t involved doing any of it for me. Instead, it’s Anthropic’s Claude Code , which I use to make software that streamlines all the work that surrounds my real work. Along with my word processor, I’ve built myself a note-taker, an email client, a Bluesky-Mastodon-Threads crossposter, and an RSS reader. They’re exactly the apps I always wanted. And every minute they save me is more time I can invest in conducting interviews and other research, sussing out what they tell me, and pounding out copy. I’m not arguing that my approach is the only acceptable one. Alex Heath, founder of the Sources newsletter , told Maxwell Zeff of Wired that he focuses on snagging scoops and leans heavily on AI for the writing aspect of his work. I appreciate the disclosure, and—more importantly—am impressed by the results. May the day come when everyone in this business figures out how to use the technology in ways that will never leave them feeling sheepish, or worse. You’ve been reading Plugged In , Fast Company ’s weekly tech newsletter from me, global technology editor Harry McCracken. If a friend or colleague forwarded this edition to you—or if you’re reading it on fastcompany.com—you can check out previous issues and sign up to get it yourself every Friday morning. I love hearing from you: Ping me at hmccracken@fastcompany.com with your feedback and ideas for future newsletters. I’m also on Bluesky , Mastodon , and Threads , and you can follow Plugged In on Flipboard. More top tech stories from Fast Company SpaceX’s $1.75 trillion IPO pitch relies on a lot of AI faith The rocket launching company says it’ll become a major player in enterprise AI in the future. Read More → Microsoft’s AI Copilot is getting a human-focused streamlining The company is aiming to make the software easier to use for a growing number of workplace and personal tasks. Read More → Why are big AI companies embedding engineers with customers, and what does that mean? If intelligence were a true utility, you wouldn’t need to send people to every customer to make the faucet work. Read More → Oura shrunk its new ring by 40% and made it look a lot more like jewelry Customers wanted a smart ring that looked like jewelry, not a gadget. Oura spent years rebuilding its hardware from the inside out to deliver one. Read More → Eileen Collins was the first woman to command a space shuttle. A new documentary shows how she got there ‘Spacewoman’ chronicles how Collins persevered against odds to become the first female space shuttle pilot and commander. Read More → The AI backlash is a danger for every brand now Brands are particularly vulnerable to charges of inauthenticity, and AI is currently an inauthenticity force multiplier. Read More →
- Yes, Schools Should Allow Students to Use AI (Opinion)
A high school student defends AI use in this letter to the editor.
Score: 35🌐 MovesMay 29, 2026https://www.edweek.org/technology/opinion-yes-schools-should-allow-students-to-use-ai/2026/05 - A.I. Hallucinations on the Hill
A.I. Hallucinations on the Hill Puck
- AD Ports, NYU Abu Dhabi to develop AI system for smarter port operations
AD Ports, NYU Abu Dhabi to develop AI system for smarter port operations Gulf News
- Latin American Unit of Top-Ten Global Life Insurer Selects SalesCloser for AI-Driven Customer Engagement and Onboarding
Latin American Unit of Top-Ten Global Life Insurer Selects SalesCloser for AI-Driven Customer Engagement and Onboarding Toronto Star
- AI shopping companions and the talent reset in retail
The pantry on the eighth floor was unusually quiet that morning. Several employees sat with coffee cups in their hands while large dashboards displayed customer behaviour, inventory movement, and real-time promotion analytics. Yet the discussion inside the room was not about sales targets or product shortages. It was about something bigger. Talent reset. “AI is […] The post AI shopping companions and the talent reset in retail appeared first on e27 .
Score: 35🌐 MovesMay 29, 2026https://e27.co/ai-shopping-companions-and-the-talent-reset-in-retail-20260527/ - How to build multi-agent systems with MCP
Assign a task with a single focus to an AI agent, and it'll get the job done without a hitch. But ask it to handle work that spans multiple tools, formats, and decisions, and that same agent could hit a wall, its reasoning degrading under the weight of your request. Multi-agent systems fix that by splitting up the work. Instead of one agent doing everything, you've got a team of them, each handling a smaller piece of the job. The tricky part is getting each agent to share tools, pass outputs, an
- Why enterprise AI fails without a strong data foundation
By Rohit Kumar, COO, SCIKIQ Almost every large company wants AI now. The board has asked, the CEO read something on a flight, and there is usually money set aside […] The post Why enterprise AI fails without a strong data foundation appeared first on Express Computer .
Score: 35🌐 MovesMay 29, 2026https://www.expresscomputer.in/guest-blogs/why-enterprise-ai-fails-without-a-strong-data-foundation/135527/ - Oslo-based Cloudgeni raises €858k to build reliable AI agents for secure cloud infrastructure
Cloudgeni, an Oslo-based startup developing specialised AI agents that build and operate cloud infrastructure in a safe way, has raised €858k ($1 million) in fresh funding to scale further across the Nordics and the US. The funding was secured from a group of Nordic investors, including the byFounders Angel Collective, an angel investor network affiliated […] The post Oslo-based Cloudgeni raises €858k to build reliable AI agents for secure cloud infrastructure appeared first on EU-Startups .
- How to Orchestrate AI Agents Across Your SDLC with Jira
How to Orchestrate AI Agents Across Your SDLC with Jira Atlassian
Score: 35🌐 MovesMay 29, 2026https://www.atlassian.com/webinars/enterprise-cloud/put-ai-agents-to-work-across-your-sdlc - Five Approaches CIOs Can Use to Build Winning Cases for AI Investment
Five Approaches CIOs Can Use to Build Winning Cases for AI Investment Gartner
- How Braintrust turns customer requests into code with Codex
How Braintrust engineers use Codex with GPT-5.5 to run experiments and code faster.
- AI Slop Is Coming for Your Playlists
Is that song stuck in your head actually AI?
Score: 35🌐 MovesMay 29, 2026https://www.theatlantic.com/newsletters/2026/05/ai-slop-music/687359/?utm_source=feed - RAG Is Burning Money — I Built a Cost Control Layer to Fix It
Most RAG systems are optimized for answer quality, not cost—and that blind spot gets expensive fast. In this article, I break down a production-ready cost control layer combining semantic caching, query routing, token budgeting, and circuit breaking, achieving an 85% reduction in LLM costs without sacrificing answer quality. The post RAG Is Burning Money — I Built a Cost Control Layer to Fix It appeared first on Towards Data Science .
Score: 35🌐 MovesMay 29, 2026https://towardsdatascience.com/rag-is-burning-money-i-built-a-cost-control-layer-to-fix-it/ - The next AI fight: Do the chatbots have First Amendment rights?
The standoff between Anthropic and the Pentagon is even bigger than it looks. And it’s even stranger than it seems
Score: 35🌐 MovesMay 29, 2026https://qz.com/ai-first-amendment-rights-anthropic-claude-pentagon-regulation-law - What would AI with constitutional rights actually look like?
The legal building blocks for AI personhood already exist, scattered across corporate law, animal rights cases, and First Amendment doctrine
Score: 35🌐 MovesMay 29, 2026https://qz.com/ai-constitutional-rights-legal-personhood-implications-052726 - Virtual AI testbed lets developers verify massive LLM servers before construction
Operating large language model (LLM) services like ChatGPT requires a server infrastructure on the scale of tens of thousands of units. However, constructing actual equipment every time a new AI semiconductor or system architecture needs to be verified incurs massive costs and time.
- Google is loosening Gemini's new usage limits after users hit caps too fast
Google is capping how much quota a single prompt can use and making Flash-Lite prompts free after users burned through limits rapidly
Score: 34🌐 MovesMay 29, 2026https://qz.com/google-gemini-usage-limits-loosening-subscriber-backlash-052926 - A University of Florida Professor Stopped Fighting AI in His Classroom: A Peer-Reviewed Study Followed
When Dr. Brian Harfe first noticed that AI tools could answer his essay prompts better than most of his students, he did not panic. He redesigned the assignment. That decision, made in a large-enrollment course at the University of Florida, has since become the subject of a peer-reviewed study published in TechTrends in 2026. What […] The post A University of Florida Professor Stopped Fighting AI in His Classroom: A Peer-Reviewed Study Followed appeared first on Grammarly Blog .
Score: 34🌐 MovesMay 29, 2026https://www.grammarly.com/blog/institutions/ai-writing-assignments-research/ - AI and ultralow-energy lasers enable an ultrafast authentication system
The security of modern communications heavily relies on systems that can rapidly and reliably verify users and the devices they are using. This process, known as authentication, essentially entails confirming that users or devices are legitimate (i.e., who or what they claim to be).
Score: 34🌐 MovesMay 29, 2026https://techxplore.com/news/2026-05-ai-ultralow-energy-lasers-enable.html - YouTube's AI Will Let You Customize Your Home Feed
YouTube's new AI features could be a big deal for content creators.
Score: 34🌐 MovesMay 29, 2026https://www.cnet.com/tech/services-and-software/youtube-rolls-out-new-ai-prompted-custom-feeds/ - Jim Cramer says Dell’s blowout quarter sets up a crucial week for AI stocks
CNBC’s Jim Cramer said Dell Technologies’ blockbuster quarter reignited enthusiasm around AI and data center stocks.
Score: 33🌐 MovesMay 29, 2026https://www.cnbc.com/2026/05/29/jim-cramer-says-dells-blowout-quarter-sets-up-a-crucial-week-for-ai-stocks.html - AI Is Changing How We Write Infrastructure, But It’s Not Solving How We Control It
AI Is Changing How We Write Infrastructure, But It’s Not Solving How We Control It DevOps.com
Score: 33🌐 MovesMay 29, 2026https://devops.com/ai-is-changing-how-we-write-infrastructure-but-its-not-solving-how-we-control-it/ - BUV featured in global report on AI governance in higher education
British University Vietnam (BUV) has been featured in a global report by Pearson and Amazon Web Services (AWS) as a case study for its approach to AI governance in higher education through the university's AI Assessment Scale framework.
- The future of product craft: Why AI-native PMs build better products, not just work faster
The future of product craft: Why AI-native PMs build better products, not just work faster Atlassian
- Cognition’s Scott Wu says AI coding agents shouldn’t replace humans
Cognition makes Devin, the first and arguably most successful AI coding agent. But famed coder Wu says it isn't designed to supplant human programmers.
Score: 33🌐 MovesMay 29, 2026https://techcrunch.com/2026/05/29/cognitions-scott-wu-says-ai-coding-agents-shouldnt-replace-humans/ - The Uncomfortable Truth AI Is Forcing Every Business to Face
The Uncomfortable Truth AI Is Forcing Every Business to Face entrepreneur.com
- Ministry adjusts focus to second phase of AI project
The Digital Economy and Society (DES) Ministry plans to move ahead with development of the second phase of the TH-AI Passport project at a cost of 900 million baht, despite ongoing criticism of the first phase.
Score: 32🌐 MovesMay 29, 2026https://www.bangkokpost.com/business/general/3262915/ministry-adjusts-focus-to-second-phase-of-ai-project - The Biggest Tell That Something Was Written by AI
Look closely and you’ll see that every part of the text is not quite right.
Score: 32🌐 MovesMay 29, 2026https://www.theatlantic.com/technology/2026/05/how-to-tell-ai-writing/687345/?utm_source=feed - How to Create a SaaS App with AI in 7 Easy Steps (2026 Guide)
Build your own SaaS app for work, productivity or community with our step-by-step guide. Bubble’s AI will help you generate the foundation, then we’ll walk you through how to customize and complete your app.
- Building workflows for agents with Skills and Interpreters
Building workflows for agents with Skills and Interpreters
- Can a Rubric Gate Stop an Agent From Taking the Wrong Action?
Inspired by Claude Outcomes, I tested a small outcome-gated retry loop on 30 support decisions. Wrong final actions dropped from 6 out of 30 to 2 out of 30, but the remaining failures showed why detection is not the same as repair. Baseline Path vs Gated Path with one retry. Source: Created by the author using OpenAI Images 2.0 Anthropic’s Claude Outcomes feature caught my attention because it formalizes a pattern many teams are already building by hand: a worker agent produces an answer, a separate grader checks the answer against a rubric, and the agent gets another pass when the output misses the mark. I liked the direction, but I wanted to test the smaller pattern underneath it without depending on the managed-agent product itself. I rebuilt the basic loop in a controlled setup: agent output, rubric judge, and one retry. I narrowed the experiment to one practical failure mode: Can a rubric-gated retry loop reduce wrong final actions when an agent has to make a structured decision? I focused on final actions because many production agent failures are not writing failures. The answer can sound reasonable, the explanation can be fluent, and the JSON can be valid, while the selected action is still wrong. In a support workflow, that becomes an operational mistake, not just a response-quality issue. An agent that says DENY when it should say NO_ACTION is still wrong, even if the reason sounds cautious. So I built a small experiment around that failure mode, using 30 synthetic support cases and a single model, gpt-5.4-mini. The sample is small, so I would not call this a benchmark. The useful part is the shape of the failures: what the gate fixed, what it missed, and what that tells you about where this pattern is worth using. What I built Each case gives the agent a short customer scenario, a policy, and a closed set of possible final actions: APPROVE DENY ASK_CLARIFYING_QUESTION ESCALATE NO_ACTION I compared two loops. The baseline sends the case straight to the agent and records whatever action comes back. The gated loop adds a rubric judge after the agent’s first answer. If the judge fails the answer, the agent gets one retry with the failed criteria. Before getting into the results, this is the pattern I would use as a quick reference for outcome-gated agent workflows. A quick guide for outcome-gated agent workflows, showing how rubric gates check proposed actions against policy contracts before accepting a final decision. Source: Created by the author using OpenAI Images 2.0 In this experiment, I kept the contract intentionally narrow: one closed action set, explicit policy constraints, and one retry after a failed rubric check. Baseline: case -> agent -> final action Gated: case -> agent -> rubric judge if failed -> retry once retry answer -> final action The cases were synthetic but designed around common production-style constraints: I used a real LLM provider rather than a deterministic simulator because the interesting part is how the model actually behaves when its first answer gets judged and revised. The important design choice The judge should not simply know the answer key. Each case has an expected_action, but that field is used only for offline metrics after the run. The judge never sees it. Instead, the judge receives the agent's selected action and checks whether that action violates objective policy constraints. For example, a case may say that a refund request already has an open chargeback in the payments workflow. The expected action may be NO_ACTION, because support should not decide the case while payments owns the resolution. A weak judge would check: “Expected action is NO_ACTION. Did the model choose NO_ACTION?” That is an answer-key check. It is useful for offline scoring, but it is not a rubric gate. The rubric gate asks something closer to: “If the selected action is DENY, does it incorrectly make support the decision owner while another workflow owns the case?” In production, you usually do not want the runtime judge to be a hidden answer key. You want it to check whether the agent’s proposed action violates a clear contract. In this experiment, the contract was simple: choose one final action, and do not choose an action that conflicts with the policy constraints in the case. The result In one run across 30 cases, the baseline agent made 6 wrong final-action decisions. After adding the rubric judge and one retry, wrong final actions dropped to 2. The improvement was meaningful, but the table is not the whole story. The gated loop did not make the agent generally smarter. It helped only when the judge could point to a specific mismatch between the selected action and the policy contract. That distinction became clearer in the failure examples. What the gate fixed The gated loop fixed 4 of the 6 baseline failures. Most of the fixes came from cases where the baseline agent was too quick to deny the request. The policy did not support immediate approval, but it also did not require denial. The better action was to ask for clarification or wait for another workflow. Here is one example. A requester asked support to change the account email. The request came from a new device, and the requester had not completed MFA. The requester did know the billing ZIP code. The policy said that account email changes require completed MFA or another approved ownership check before support can make the change. The baseline chose DENY. The expected action was ASK_CLARIFYING_QUESTION. The baseline answer was too final. The requester still had a possible path to complete an approved ownership check, so the right next step was to ask about that, not to close the door. The rubric judge failed the answer because denial was premature. After retry, the model changed the action to ASK_CLARIFYING_QUESTION. Another fixed case involved a replacement device. The customer asked support to resend a replacement, but an RMA had already been created that morning and was waiting for warehouse scan. The baseline chose DENY. The expected action was NO_ACTION. Denying the request misrepresented the workflow. The issue was not that the customer was ineligible. The issue was that the replacement process was already underway and support should not create a duplicate RMA. The retry corrected this to NO_ACTION. These examples show the value of the pattern at a practical level. The gate did not make the model smarter in a broad sense. It forced the model to repair a specific mismatch between its selected action and the policy contract. What still failed The two remaining failures were more interesting than the aggregate score. Both were existing workflow cases. In both, the expected action was NO_ACTION, but the final gated answer remained DENY even after retry. The first case involved a refund request with an open chargeback in the payments workflow. The policy said support must not issue refunds while a chargeback workflow is active, because the payment workflow owns the resolution. The expected action was NO_ACTION. The final action after gating was DENY. The second case involved a buyer asking for manual shipment release while a fraud review workflow was still open. The policy said support must not release shipments while fraud review is open. Again, the expected action was NO_ACTION. The final action after gating was DENY. In both cases, the judge identified the issue and returned failed criteria. The retry still kept the wrong action. This was the part that changed how I would think about the pattern in production. A rubric gate can detect that an action violates a rule, but detection is not the same as repair. If the model does not understand the operational distinction between “deny the request” and “take no action because another workflow owns the decision,” the retry can preserve the wrong action even after the judge flags it. A safe-sounding action is not always the correct action. DENY may sound conservative, but it is still wrong if the agent has no authority to decide the case. Why objective outcomes matter This pattern worked because the outcome was checkable. The agent had to choose from a closed action set, the policy constraints were explicit, and the judge could inspect the selected action and ask whether it violated a specific rule. For production systems, I would trust outcome-gated retries more in cases like these: I would be more cautious when the outcome is subjective: What I would change in production I would not use this pattern as a universal reliability layer. I would use it where the cost of the wrong action is high enough to justify an extra model call, and where the expected result can be expressed as a contract. The most important production change would be treating repeated non-repair cases as design feedback. If the agent keeps confusing DENY and NO_ACTION, that is a signal to improve the action definitions, add examples, or restructure the workflow contract. Another retry is unlikely to fix a conceptual gap in the model's understanding. Beyond that, I would separate runtime gating from offline scoring. The runtime judge checks constraints. The offline evaluator compares final actions against labeled expected outcomes. Mixing those two roles makes the system look stronger than it is. I would also log every retry with the original action, failed criteria, revised action, and final pass/fail status, because the pattern of repairs and non-repairs is often more useful than the final accuracy number. Closing Outcome gates are most useful when “done” can be written as a contract. In this experiment, that contract was simple: choose one final action that does not violate the policy constraints. Under those conditions, the gated loop reduced wrong final actions from 6 out of 30 to 2 out of 30. The remaining failures are the part I would pay most attention to in production. The judge detected the problem, but the retry still returned the wrong final action. That means the reliability question is not only whether a system can grade an output. It is whether the full loop can turn that feedback into the right decision. That is the part I would want evidence for before putting an outcome-gated loop into a real workflow. The repo for this experiment includes the synthetic cases, baseline runner, gated runner, judge logic, and generated result files from this run. References Anthropic Claude API Docs: Define outcomes Anthropic Cookbook: Outcomes, agents that verify their own work GitHub repo for this experiment Can a Rubric Gate Stop an Agent From Taking the Wrong Action? was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.
- The First Class of AI Natives Is Graduating. Offices Are Getting Ready.
They face cuts to entry-level jobs. They’re also highly sought after for their AI skills.
Score: 32🌐 MovesMay 29, 2026https://www.wsj.com/tech/ai/ai-natives-graduates-job-cuts-6bab8ac9?mod=rss_Technology - By the Numbers: What the class of 2026 job market actually looks like — and where AI fits in
Here's the good, the bad and the AI.
- Honolulu planning office uses ‘TurboTax-like’ AI tool to help applicants reduce mistakes
The planning agency inside the city and county government of Honolulu is using AI tools to reduce delays and improve application quality.
Score: 32🌐 MovesMay 29, 2026https://statescoop.com/honolulu-planning-office-ai-tool-reduce-mistakes/