The Vergecast 2025 Year in Review
2025 was defined by an AI arms race among OpenAI, Google, Anthropic and other rivals, with Nvidia soaring to become the world’s most valuable company. Data centers multiplied globally as every app, device and enterprise pursued AI strategies. The Vergecast’s year‑in‑review podcast recapped these trends, highlighting the best and worst tech moments and inviting listener participation. The episode also previewed bold predictions for 2026 despite past misses.
How Nabeel Ehsan Is Re-Engineering Procurement Governance with AI Tools to Detect Manipulation and Collusion?
Nabeel Ehsan is reshaping procurement governance by deploying AI‑driven forensic analytics to spot manipulation and collusion in real time. At Mawarid Holding, his data‑enabled risk framework has recovered over AED 50 million and generated AED 45 million in savings through automated controls. He oversees...
Microsoft Pledges $17.5 Billion to Expand AI, Cloud Capacity in India
Microsoft announced a $17.5 billion four‑year plan to expand AI infrastructure and cloud capacity in India, adding to a $3 billion commitment made earlier this year. The funding will power new data‑centre zones, sovereign‑cloud offerings, and a programme to skill 20 million Indians...
Ads Aren’t Coming to Gemini, Google Says — but OpenAI Has Shown that the Inevitable Is Coming
OpenAI recently introduced ad-like “app suggestions” in its ChatGPT Plus service, prompting a swift rollback after user criticism. Google’s Gemini, its chief rival, publicly denied any plans to embed advertisements, emphasizing its current ad‑free experience. Industry analysts, however, argue that...
The Download: A Peek at AI’s Future
The latest edition of MIT Technology Review’s "The Download" spotlights a heated debate over AI’s trajectory, contrasting forecasts that generative AI could eclipse the Industrial Revolution with a more measured view that adoption follows human‑speed cycles. It marks the final...
IFixit’s FixBot Helps with Repairs ‘the Way a Master Technician Would’
DIY repair platform iFixit has released a new mobile app for iOS and Android that bundles its extensive repair guide library with a battery‑health monitor and an AI chatbot called FixBot. FixBot draws on iFixit’s guides, PDF manuals and user...
India Proposes Sweeping AI–Copyright Overhaul with ‘One Nation, One Licence, One Payment’ Model
India's Department for Promotion of Industry and Internal Trade (DPIIT) has drafted a sweeping policy that mandates a compulsory, blanket licence for AI developers to use any lawfully accessed copyrighted content. Under the "one nation, one licence, one payment" model,...
AGI Is a Pipe Dream Until We Solve One Big Problem, AI Experts Say, Even as Google Celebrates Gemini's Success
At NeurIPS 2025 AI researchers warned that the current scaling‑centric approach to large language models has hit a "scaling wall," delivering only marginal gains despite massive data and compute. Google’s Gemini 3 showcased impressive performance through architectural tweaks, yet it remains...
DoJ Takes Down Sophisticated Network Smuggling $160m Worth of AI Chips to China
The U.S. Department of Justice dismantled a sophisticated smuggling ring that moved roughly $160 million worth of Nvidia H100 and H200 AI GPUs to China, breaching export restrictions. Operation “Gatekeeper” revealed the chips were stripped of Nvidia branding and relabeled as...
EcogenX Innovations and ST Engineering Win for Sustainable Solutions at the Mint All About AI Tech4Good Awards
EcogenX Innovations and Singapore Technologies Engineering (ST Engineering) captured the Gold and Silver awards respectively in the Mint All About AI Tech4Good Awards for Best Use of AI for Sustainability. EcogenX’s AI‑driven waste‑management platform, featuring smart bins, sensors and analytics,...
EU Opens Probe Into Google for Training AI without Paying Content Creators
The European Commission has opened an antitrust investigation into Google, alleging the company used YouTube videos and other online publisher content to train its generative‑AI models without paying creators or offering opt‑out options. The probe focuses on potential unfair terms...
Rezo.ai and Nivinsure Honoured for Financial Inclusion at the Mint All About AI Tech4Good Awards
Rezo.ai and Nivinsure were recognized at the Mint All About AI Tech4Good Awards for pioneering AI solutions that advance financial inclusion in India. Rezo.ai earned the Gold award for its Agentic AI platform that combines OCR, RPA and multilingual support...
US to Allow Nvidia H200 Chip Shipments to China, President Trump Says
The U.S. government announced that Nvidia’s H200 AI processors will be allowed for export to China, subject to a 25% fee collected through Taiwan. The fee is higher than the 15% rate previously discussed and applies to other AI‑chip firms...
Brand-Context AI: The Missing Requirement for Marketing AI
Marketing teams are adopting generative AI, but outputs often miss brand, audience, and strategic alignment because models lack contextual intelligence. BlueOcean argues that the missing ingredient is structured brand‑context, which unifies vertical data streams into a horizontal view for decision‑quality...
OpenAI Races Gemini 3 to the Top with GPT‑5.2 Drop This Week
OpenAI announced a “code red” and is accelerating the rollout of GPT‑5.2 to counter Google’s Gemini 3. The new model, expected by Dec 9, promises faster response times, stronger reasoning and fewer glitches, emphasizing core performance over new features. Gemini 3 has been...
Z.ai Debuts Open Source GLM-4.6V, a Native Tool-Calling Vision Model for Multimodal Reasoning
Zhipu AI’s Z.ai has launched the GLM-4.6V series, an open‑source vision‑language model family featuring a 106‑billion‑parameter flagship and a 9‑billion‑parameter Flash variant. Both models introduce native multimodal function calling, allowing visual inputs to be passed directly to tools such as...
US to Allow Exports of Advanced Artificial Intelligence Chips to China
President Donald Trump announced an agreement with Chinese President Xi Jinping that will allow Nvidia to export its H200 artificial‑intelligence GPUs to approved Chinese customers, reversing a key Biden‑era restriction. The deal excludes Nvidia’s most advanced Blackwell and upcoming Rubin...
Department of Commerce Approves Nvidia H200 Chip Exports to China
The U.S. Department of Commerce has approved export licenses for Nvidia’s H200 AI accelerator to be shipped to Chinese customers, despite a newly introduced congressional bill seeking to block such sales. The H200, built on Nvidia’s Hopper architecture, offers top‑tier...
'They Can Build a Hospital in a Weekend': Nvidia CEO Warns About China's Big AI Advantages — as Report Reveals...
Nvidia CEO Jensen Huang warned that China can construct AI‑critical data centers and even hospitals in a weekend, highlighting a stark speed advantage over the United States, where similar projects take years. A recent report shows China now accounts for...
Anthropic Is Bringing Claude Code to Slack
Anthropic has rolled out Claude Code inside Slack as a beta research preview, letting users tag Claude on coding‑related messages to invoke the AI directly. The integration pulls context from the Slack thread and any authenticated code repositories, eliminating the...
Amazon’s “Catalog AI” Product Platform Helps You Shop Smarter
Amazon launched Catalog AI in July, an automated system that harvests product information from the web and uses large language models to enrich and standardize listings. The platform builds on a glossary created by engineering leader Abhishek Agrawal, ensuring consistent...
You Can Buy Your Instacart Groceries without Leaving ChatGPT
Shoppers will be able to go from recipe planning to checkout within the ChatGPT window.
Robot Reassembles Pompeii Artifacts
A University of Bonn team equipped a dual‑arm robot with AI‑driven vision to reassemble shattered fresco fragments from Pompeii, converting broken pieces into digital scans and using a model to determine correct fits. The system, developed under the EU‑funded RePAIR...
‘ONE RULE’: Trump Says He’ll Sign an Executive Order Blocking State AI Laws Despite Bipartisan Pushback
Former President Donald Trump announced he will issue a "ONE RULE" executive order aimed at blocking state-level artificial intelligence regulations. He argued that companies should not need to secure dozens of approvals for each AI initiative. The proposal comes amid...
Proactive Hearing Assistant Filters Through Voices in a Crowd
Researchers at the University of Washington unveiled a proactive hearing assistant that isolates and amplifies only the voices of conversational partners in noisy settings. The system relies on AI‑driven turn‑taking detection rather than direction, loudness, or proximity, and operates with...
Data Centers Are Throttling Home Building as Infrastructure Struggles to Keep up with AI Demands
Data‑centre expansion driven by AI and cloud workloads is straining electricity grids across the UK and Europe. In Britain, annual demand could jump from 10 TWh today to 71 TWh by 2050, with 100 new facilities slated, half near London. The surge...
ChatGPT Users Are Making a Spotify Wrapped for Their Year in AI — Here's How to Do It
ChatGPT users can now generate a personal "Wrapped"‑style recap of their AI interactions with a single prompt. The request, "Make me a ChatGPT Wrapped summary like a Spotify Wrapped showing my usage this year," prompts the model to analyze a...
Self-Healing IT Is No Longer Science Fiction – It’s Driving Businesses Forward
Self‑healing, agentic AI is moving from sci‑fi concept to operational reality, enabling IT systems to detect, diagnose and remediate issues autonomously. These tools continuously monitor networks, endpoints and cloud services, intervening before users notice problems. Organizations can set governance rules...
Gartner Advisory States AI Browsers Are NOT Your Friend — and They Are Putting Your Business at Risk
Gartner’s newest advisory urges organizations to block AI‑enhanced browsers, citing security gaps that prioritize user experience over data protection. Agentic browsers, such as OpenAI’s ChatGPT Atlas, can autonomously navigate sites, summarize content, and perform actions within authenticated sessions. These capabilities...
Booking.com’s Agent Strategy: Disciplined, Modular and Already Delivering 2× Accuracy
Booking.com has turned its homegrown conversational recommendation system into a disciplined, modular AI agent stack that blends small travel‑specific models with larger LLMs and in‑house evaluations. This hybrid approach has doubled accuracy on key retrieval, ranking and customer‑interaction tasks while...
Square’s Product Chief on the Death of the Penny and the Future of Money
Square’s product chief Willem Avé explained Block’s recent functional re‑org, which now lets Square share engineering talent with sister units like Tidal. The new "Square 3.0" roadmap focuses on embedding deterministic AI that turns large‑language‑model insights into reliable back‑office automation...
FBI Warns of Kidnapping Scams as Hackers Turn to AI to Provide 'Proof of Life'
The FBI has issued a public service announcement warning that criminals are leveraging generative AI to produce deep‑fake videos as false "proof of life" in kidnapping and extortion scams. Scammers harvest personal photos from social media, synthesize convincing footage, and...
Hinge Launches AI Feature so You Stop Starting the Conversation with 'Hey'
Hinge has launched Convo Starters, an AI‑driven feature that offers three personalized opening suggestions when a user likes a profile. The tips draw on the match’s prompt answers and photos, and can be toggled on or off in settings. This...
OpenAI Boasts Enterprise Win Days After Internal ‘Code Red’ on Google Threat
OpenAI announced that ChatGPT Enterprise usage surged eight‑fold year‑over‑year, with employees reporting an average productivity gain of one hour per day. The rapid growth comes amid an internal “code red” alert over a looming competitive threat from Google’s AI offerings....
AI Language Models Show Bias Against Regional German Dialects
The episode discusses a new study revealing that large language models, including GPT‑5 and open‑source alternatives, systematically assign negative stereotypes to speakers of German regional dialects compared to Standard German, influencing decisions in hiring, education, and other contexts. Researchers from...
Sarvam AI and Ridlan AI Foundation Awarded for Empowering People with Disabilities
The Mint All About AI Tech4Good Awards honored Sarvam AI with the Gold prize and Ridlan AI Foundation with Silver for using artificial intelligence to empower people with disabilities in India. Sarvam AI’s sovereign, multilingual generative‑AI platform now supports 11...
AI for Education: Niveda Foundation and Udhyam Learning Foundation Win Top Place at All About AI Tech4Good Awards
The Mint All About AI Tech4Good Awards honored Niveda Foundation with the Gold and Udhyam Learning Foundation with the Silver for Best Use of AI in Education & Livelihoods. Niveda’s AI‑driven skilling platform cut dropout rates from 40% to 12%,...
A New UK AI Says It Can Beat ChatGPT – We Tried It and Here’s What We Found
Locai, a British‑built AI from Locai Labs, claims to outperform GPT‑5, Gemini and DeepSeek on the Arena Hard v2 benchmark. The system relies on a novel Forget‑Me‑Nt training method that generates its own data and avoids catastrophic forgetting. It also...
Design in the Age of AI: How Small Businesses Are Building Big Brands Faster
Generative AI has turned design from a late‑stage expense into a front‑line capability for small businesses. Since 2022, searches for AI‑powered naming, logo and website generators have surged 700‑1,600%, indicating rapid adoption. Unified platforms like Design.com now deliver naming, logo...
Gemini’s Monthly Active User Base Grew 5x Faster than ChatGPT, New Data Shows
Google’s Gemini chatbot is closing the gap with OpenAI’s ChatGPT, posting a 30% rise in monthly active users (MAU) from August to November 2025 versus ChatGPT’s 6% growth. The surge is attributed to Gemini’s new Nano Banana image‑generation model and Google’s...
Last Week in AI #328 - DeepSeek 3.2, Mistral 3, Trainium3, Runway Gen-4.5
The episode reviews a wave of new AI models and infrastructure, highlighting DeepSeek's reasoning‑first V3.2 series with tool‑use capabilities, Mistral's multimodal Mistral 3 family that runs offline on a single GPU, AWS's Trainium 3 chips and upcoming Trainium 4 with NVLink support, and...
We May Have to Wait Until 2027 for the Launch of the Next Pair of Meta Mixed Reality Smart Glasses
Meta’s next mixed‑reality smart glasses, codenamed Phoenix, have been pushed back from a 2026 debut to 2027, according to internal memos. The delay is intended to give engineers more breathing room to perfect the user experience and hardware reliability. Phoenix...
A Very Human Vision for Going All-In on AI
Founder Sari Azout discusses how Sublime, a curation platform centered on human taste, is leveraging AI to augment rather than replace creativity. The Vergecast episode highlights Sublime’s AI‑driven discovery features and its new Podcast Magic tool, which runs entirely on...
Why Cybersecurity Is Now a Strategic Lever for Tech Firms Navigating Uncertainty
Cybersecurity has shifted from a technical afterthought to a strategic lever that directly influences enterprise value for technology firms, especially SaaS and AI players. Investors and private‑equity owners now embed security posture—SOC 2, ISO 27001, and cultural maturity—into due‑diligence and deal pricing....
Softbank’s Son Says He ‘Cried’ While Selling Nvidia Stake to Fund OpenAI Bet
SoftBank CEO Masayoshi Son disclosed that the group sold its entire $5.8 billion Nvidia stake to free cash for a $30 billion commitment to OpenAI and other AI projects. The shares were sold at $181.58, about 14 % below Nvidia’s all‑time high, prompting...
Why AI Coding Agents Aren’t Production-Ready: Brittle Context Windows, Broken Refactors, Missing Operational Awareness
AI coding agents can generate snippets quickly, but they falter in enterprise settings due to limited context windows, service limits, and lack of hardware awareness. Indexing caps at 2,500 files and 500 KB per file leave large monorepos partially invisible, forcing...
This Chinese Chip Giant Is Boosting Production to Try and Take on Nvidia - but How Will Huawei Feel?
Cambricon Technologies announced plans to produce 500,000 AI accelerator chips in 2026, roughly tripling its 2025 output. The target includes 300,000 units of its Siyuan 590 and 690 models, but current yields sit at only 20 percent. The company relies on...
AI ‘Creators’ Might Just Crash the Influencer Economy
Influencer Jeremy Carrasco warns AI‑generated videos are flooding TikTok and Instagram, threatening the creator economy. He highlights tell‑tale signs such as wobbly eyes, “Sora Noise,” and inconsistent backgrounds that expose low‑quality AI clips. Monetization of AI content can yield modest...
HTLS 2025: Can Doctors and Scientists Trust AI with Their Work? Google DeepMind's Pushmeet Kohli Answers
At the Hindustan Times Leadership Summit 2025, DeepMind VP Pushmeet Kohli addressed whether doctors and scientists can trust AI. He highlighted AlphaFold’s high accuracy paired with explicit uncertainty signals, underscoring the need to know when AI is right. Kohli warned that...
YouTube Creators Are Now Making AI Slop for Babies: Report
A wave of YouTube creators is mass‑producing low‑effort AI‑generated videos, dubbed “AI slop,” that are appearing on YouTube Kids and reaching toddlers under two. Bloomberg reports that creators with millions of followers are posting tutorials on how to make such...