Welcome to this week in AI and tech — where AI bugs make time travel possible, humanoid robots get open-source swag, and privacy meets automation at Meta. From Google fixing its AI time warp to new AI-powered tools that turn your inbox and spreadsheets into productivity playgrounds, it’s a whirlwind of innovation, controversy, and a peek at the future of tech.
Buckle up, because here’s everything you need to know from May 26–June 1, 2025 — no glitches, just pure byte-sized brilliance.
Google Fixes AI Bug That Made Overviews Say It’s 2024 Instead of 2025
What Happened?
Google’s AI Overviews, designed to help with research and answering questions, was mistakenly saying the current year was 2024, even though it’s 2025. Users noticed this error when asking what year it is, and the bug persisted for several days. Google fixed the issue late on May 29, 2025. 
Why Did This Happen?
Google didn’t explain the exact cause but said they are continuously updating and improving AI Overviews to prevent such errors. A spokesperson emphasized that most AI Overviews responses are accurate, and the company is actively working on fixes for issues like this.
What Else?
AI Overviews has had other notable blunders before, such as bizarre advice to eat rocks for vitamins and suggesting glue to help cheese stick to pizza. Its rollout in Hindi in India also showed inconsistent and confusing answers. Despite these glitches, Google CEO Sundar Pichai reported that AI Overviews is now used by 1.5 billion users across 100+ countries and drives over 10% of search queries in markets like the U.S. and India. Google is heavily investing in AI-powered search and Q&A features to boost user engagement.
Hugging Face Launches Two New Open Source Humanoid Robots: HopeJR and Reachy Mini
What’s New?
Hugging Face unveiled two new humanoid robots: HopeJR, a full-size robot with 66 independent movements capable of walking and arm motion, and Reachy Mini, a desktop robot that can move its head, talk, listen, and serve as a test platform for AI applications. Both robots are open source, allowing users to assemble, modify, and understand their inner workings.
When and How Much?
The company plans to start shipping initial units by the end of 2025, with a waitlist now open. HopeJR is expected to cost about $3,000 per unit, while Reachy Mini will be priced between $250 and $300, depending on tariffs.
Why It Matters
Hugging Face’s CEO, Clem Delangue, highlighted affordability and openness as key goals to avoid robotics being dominated by a few closed, proprietary systems. This release follows Hugging Face’s April acquisition of humanoid robotics startup Pollen Robotics, which helped provide the capabilities needed for these new bots.
Background
Hugging Face has been expanding its robotics efforts, launching the LeRobot platform in 2024—a collection of open AI models and tools for robotics development—and recently releasing an updated 3D-printed programmable robotic arm in partnership with The Robot Studio. They also expanded training data for self-driving machines through collaboration with AI startup Yaak.
Perplexity Launches Perplexity Labs to Generate Spreadsheets, Dashboards, and More
What’s New?
Perplexity, an AI-powered search engine competing with Google, released Perplexity Labs—a new tool available to subscribers of its $20/month Pro plan. Labs can create reports, spreadsheets, dashboards, interactive web apps, and other data-driven deliverables by combining web search, code execution, and visualization tools.
Where and When?
Perplexity Labs is accessible on the web, iOS, and Android, with Mac and Windows app support coming soon.
Why It Matters
Designed for deeper, longer tasks (10+ minutes), Perplexity Labs enables users to organize generated charts, images, and code in a dedicated dashboard for easy viewing and downloading. This expansion moves Perplexity beyond just search into productivity and data analysis tools.
Background
This launch coincides with similar AI-driven tools emerging in the market, like Manus’s slide deck creator. Perplexity has also been broadening its offerings with a new web browser called Comet and the acquisition of Read.vc, a professional social network. The company is increasing focus on corporate features, having introduced an enterprise plan with advanced user and knowledge management last year. Perplexity is reportedly in talks to raise up to $1 billion at an $18 billion valuation, reflecting strong investor interest.
NAACP Demands Halt of xAI’s ‘Dirty Data Center’ Operations in Memphis
What’s Going On?
The NAACP is urging Memphis officials to stop operations at Colossus, a supercomputer facility run by Elon Musk’s xAI in South Memphis, due to environmental and health concerns. They sent a letter to the Shelby County Health Department and Memphis Light Gas and Water criticizing what they call a “lackadaisical approach” to regulating the facility.
Why the Concern?
xAI operates gas turbines at Colossus that reportedly emit hazardous pollutants like formaldehyde and nitrogen oxides at levels above EPA limits. While xAI has applied for permits for 15 turbines, the NAACP alleges the company has been running at least 35 turbines without permits over the past year. The turbines’ emissions pose serious health risks, especially since the facility is near Boxtown, a historically Black community already facing cancer rates four times the national average.
What’s Next?
The NAACP demands an emergency order to stop xAI’s operations or at least enforcement action against permit violations. The letter is addressed to Shelby County Health Department Director Michelle Taylor—who is soon leaving for Baltimore—and Memphis Light Gas and Water commissioners. Memphis Light Gas and Water said they had not yet received the letter. TechCrunch reached out to both the NAACP and xAI for comment.
Meta to Automate Most Product Risk Assessments Using AI
What’s Happening?
Meta plans to use an AI-powered system to handle up to 90% of privacy and risk assessments for updates to its apps like Instagram and WhatsApp. This marks a shift from the current process, which relies heavily on human evaluators.
Why?
A 2012 agreement with the FTC requires Meta to conduct privacy reviews of product changes. The new system will have product teams complete questionnaires, after which AI will quickly generate risk evaluations and necessary conditions for updates before launch.
What Are the Implications?
This AI-driven approach aims to speed up product updates and bring more consistency. However, a former Meta executive warned NPR that it could increase risks by potentially missing negative effects before they impact users.
Meta’s Response
Meta stated it has invested over $8 billion in privacy programs and is committed to regulatory compliance. The company says AI will handle routine, low-risk decisions, while humans will continue overseeing complex or novel issues to ensure thorough assessments.
Google quietly released an app that lets you download and run AI models locally
What Just Dropped?
Google launched Google AI Edge Gallery, an Android app (iOS coming soon) that lets you find, download, and run Hugging Face AI models right on your phone — no internet needed. It runs everything offline using your phone’s processor, handling tasks like image generation, Q&A, and code editing. Plus, it features shortcuts like “Ask Image,” “AI Chat,” and a customizable “Prompt Lab” for quick AI jobs.
Why Should You Care?
This is huge for privacy and convenience. You don’t have to send your data to the cloud or worry about losing connection to get AI power. Developers get an Apache 2.0 licensed tool to experiment with and build on freely, while users get AI that works anytime, anywhere — straight from their phones.
Gemini Now Auto-Summarizes Your Long Emails — Unless You Opt Out
What Just Dropped?
Google’s AI assistant, Gemini, just got a bigger role in your Gmail inbox with new email summary cards. Instead of waiting for you to ask, Gemini will now automatically summarize long emails at the top of your messages—no clicks needed. These AI-powered summaries highlight key points and update in real-time as replies come in. The classic manual summary button isn’t going anywhere, so you can still take control if you want.
Why Should You Care?
This move shows how AI is quietly taking over everyday tools you already use, making email triage faster and less overwhelming. But heads up—auto summaries aren’t perfect yet. Google’s AI, like other tech giants’ attempts, can sometimes miss the mark or get details wrong. Still, if you deal with long email threads, Gemini’s summaries could save you time and help you stay on top of important info without digging through endless replies. You can also opt out or adjust the settings if you want to keep it old-school.
Odyssey’s AI Streams Interactive 3D Worlds You Can Explore Like a Game
What Just Dropped?
Odyssey, a startup founded by self-driving pioneers, unveiled a new AI model that streams 3D-like interactive video worlds at lightning speed — generating new video frames every 40 milliseconds. Available as an early web demo, users can explore these video spaces with simple controls, much like navigating a 3D game. Powered by a “world model,” Odyssey’s AI predicts what happens next in a scene, maintaining spatial consistency and generating coherent video streams for minutes at a time.
Why Should You Care?
This tech could revolutionize how we experience entertainment, training, ads, and education by turning passive videos into fully interactive environments on demand — no costly traditional production needed. While the demo is still rough (blurry visuals and shifting layouts), Odyssey plans rapid improvements and offers creators ways to edit AI-generated scenes with popular tools like Unreal Engine and Blender. Backed by big investors and Pixar co-founder Ed Catmull, Odyssey is aiming to reshape video into an immersive, interactive future — all powered by AI streaming.
And that’s a wrap on this week’s AI and tech headlines — from glitchy overviews and affordable humanoid robots to AI that reads your emails and Meta’s new privacy watchdog. As AI keeps breaking new ground and stirring up debate, one thing’s certain: the future is arriving faster than ever, and it’s anything but predictable.
Stay curious, stay critical, and most importantly, keep your sense of humor — because in the fast-moving world of AI, the next curveball is always just a click away.
