- Everyday AI
- Posts
- NVIDIA’s big AI advancements, Claude gets the internet, ChatGPT gets new voice models, Gemini goes Canvas and more
NVIDIA’s big AI advancements, Claude gets the internet, ChatGPT gets new voice models, Gemini goes Canvas and more
OpenAI shakes up its leadership, Microsoft unveils new AI agents for cyber, new LLM test stumps top models.
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Sup y’all! 👋
We’ve got a TON of fresh interviews coming at ya over the next two weeks from NVIDIA’s GTC conference.
Stoked for those.
But while we were out, there were a TON of new AI updates and products. What do you want us to cover in an upcoming podcast/livestream?
What do you want us to cover in an upcoming podcast/livestream? |
✌️
(Let’s connect on LinkedIn. Tell me you’re from the newsletter!)
Today in Everyday AI
8 minute read
🎙 Daily Podcast Episode: ChatGPT, Google Gemini, Anthropic Claude all drop useful updates. Give it a listen.
🕵️♂️ Fresh Finds: DeepSeek's sneaky V3 update, a $50 million AI "gift" and Google updates its new AI mode. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: OpenAI shakes up its leadership, Microsoft unveils new AI agents for cyber, new LLM test stumps top models. Read on for Byte Sized News.
🧠 AI News That Matters: Don’t waste hours a day keeping up with AI news. Each Monday, we give you the 101. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We talked about how AI is catching up with human intelligence, Yahoo hoping for an AI-powered comeback and Oracle releases AI agents.and more. Check it here!
AI News That Matters - March 24, 2025 📰
You prolly missed a lot of this AI news: 👇
Gemini got Canvas 🎨
ChatGPT got a new TTS model 🗣️
Claude finally got the internet 🕸️
NVIDIA re-invented AI computing 💻
And that's barely the surface of this week's BIG AI news.
Don't waste hours a day worry about AI.
We'll cut through the nonsense and tell you what it all means during our weekly AI News that Matters segment.
We break it down in this week’s The AI News That Matters.
Join the conversation and ask Jordan any questions on AI here.
Also on the pod today:
• ChatGPT goes TTS 🗣️
• Google drops impressive Gemini update ✨
• Claude FINALLY gets internet 📱
It’ll be worth your 57 minutes:
Listen on our site:
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Base44 uses AI to turn your ideas into products, LiftmyCV uses AI to apply to jobs for you, Pocketlink is a new, AI-powered, Link-in-Bio tool.
DeepSeek — DeepSeek quietly shipped an update to their popular V3 model, specifically making it able to run on Apple’s new Mac Studio M3 Ultra.
AI For Humanity – Netflix cofounder Reed Hastings ’83 is funding Bowdoin’s bold new AI initiative with a $50 million gift to prepare students for a world transformed by artificial intelligence.
Google's AI – Google has refreshed its AI mode, as well as rolling it out to more paid users in the U.S.
1/ Today we’re expanding access to AI Mode beyond Google One AI Premium subscribers in the US, as we begin taking people off the Labs waitlist. We’re hearing early users really appreciate the ability to get AI responses for tougher questions, along with fresh, high quality info
— Robby Stein (@rmstein)
11:33 PM • Mar 24, 2025
AI in Retail — 76% of retailers are boosting AI investments to streamline operations and enhance shopping experiences
AI Models — Alibaba's new AI model, Qwen2.5-VL-32B-Instruct, is raising the bar in math reasoning and multimodal tasks.
Big Tech Moves — Ex-Intel CEO Pat Gelsinger joins Gloo to lead AI-driven tools for faith groups, combining tech innovation with his lifelong Christian mission
1. OpenAI Shakes Up Leadership to Propel AI Innovation 🪇
OpenAI is sharpening its focus on innovation as CEO Sam Altman pivots toward research and product development, according to Bloomberg. COO Brad Lightcap will now oversee global operations and partnerships, including collaborations with Microsoft and Apple.
Meanwhile, Mark Chen and Julia Villagra step into new roles as chief research officer and chief people officer, signaling a strategic leadership expansion.
2. Microsoft Unveils AI Agents to Tackle Cybersecurity Chaos 🛡️
Microsoft has raised the stakes in cybersecurity with the announcement of new autonomous AI agents embedded within Security Copilot, set to debut this April.
These agents will tackle critical issues like phishing, data loss prevention, and identity management, addressing the overwhelming surge of cyberattacks—such as the 30 billion phishing emails detected last year. According to Microsoft, these tools aim to free up security teams by handling high-volume tasks autonomously, while also providing advanced defenses against emerging threats like "shadow AI" and indirect prompt injection attacks.
3. FuriosaAI Rejects Meta’s $800M Offer, Stays Independent 💾
South Korean AI chipmaker FuriosaAI has reportedly turned down an $800 million acquisition bid from Meta, as per a local media report. The startup, known for its Warboy and Renegade (RNGD) chips, chose to focus on its own growth rather than compromise on its business strategy and structure.
Meta’s offer highlights its urgency to reduce reliance on Nvidia amid its massive $65 billion AI push, but FuriosaAI has other plans—like raising $48 million in fresh funding this month. With RNGD chips set to launch this year, this rejection signals FuriosaAI’s ambition to challenge industry giants like Nvidia and AMD head-on.
4. Apple Finally Joins the AI Data Center Race 🖥️
Apple is upping its AI game with a $1 billion investment in Nvidia's high-powered server systems, according to reports.
Partnering with Dell and Super Micro Computer, Apple plans to build a large server cluster to support advanced AI models, potentially catching up to competitors like Microsoft and Google. This pivot comes after delays in launching its AI-enhanced Siri assistant—a move described internally as "ugly" and "embarrassing," per Bloomberg.
5. AI’s Toughest Test Yet? ARC-AGI-2 Stumps Leading Models 🧑🏫
The Arc Prize Foundation, co-founded by AI expert François Chollet, just unveiled ARC-AGI-2, a new intelligence test that has left top AI models scratching their digital heads.
According to the foundation, reasoning models like OpenAI’s o1-pro and DeepSeek’s R1 scored between 1% and 1.3%, while non-reasoning models like GPT-4.5 and Gemini 2.0 Flash barely hit 1%. Designed to challenge models with unseen visual puzzles and prioritize efficiency over brute computational power, ARC-AGI-2 aims to measure true adaptability, unlike its predecessor. With humans averaging a 60% success rate, this test highlights the gap between AI hype and reality.
NVIDIA just dropped AI desktops that'll make your gaming rig look like a calculator, while OpenAI's Sora went UNLIMITED for paying users.
Meanwhile, Pennsylvania's government saved 8 hours weekly using ChatGPT, and Apple's cooking up CAMERAS for your AirPods.
The future's weird, and it's happening NOW.
Don’t waste hours a day trying to keep up with AI.
We do that for you – not only each day, but with our Monday breakdown of the AI News that Matters.
Here’s what ya need to know to get ahead, shorties. 👇
1. NVIDIA Unveils AI Desktop Systems That'll Make Your Gaming Rig Cry 🖥️
NVIDIA just dropped two new AI-focused desktop systems at their GTC conference that'll make your gaming setup look like a potato.
The DGX Spark and DGX Station aren't playing around.
Sheeeeeesh.
These beasts are designed specifically for AI development and deployment right on your desk.
No more cloud dependencies or praying your connection stays stable.
Local AI compute is about to get SERIOUS with these new machines making previously server-only tasks possible right in your office.
NVIDIA continues their strategy of building the picks and shovels for the AI gold rush. (And like…. the trains, railways, gold refineries, gold bars…..)
What it means:
The era of cloud-only AI could by slowly dying. Companies that invested everything in cloud infrastructure could be scrambling to explain why their strategies suddenly look dated.
Local AI processing solves the two biggest enterprise AI problems: data privacy and consistent performance. This is a game-changer for sensitive industries like healthcare and finance.
By 2027, having dedicated AI compute in-house will be as standard as having servers was in 2010. Companies without this capability will be at a massive competitive disadvantage.
What a $3,000 DGX Spark provides in terms of edge AI would have cost about $40K three years ago for the equivalent.
2. NVIDIA's GPU Roadmap Through 2028 Shows They're Playing 5D Chess 🧠
NVIDIA revealed its GPU roadmap stretching all the way to 2028, and it's WILD.
The Rubin Ultra is coming in 2027, promising performance that'll make today's H100s look like pocket calculators.
The Feynman GPU series lands in 2028, named after the legendary physicist and likely to break physics in its own way.
This long-term roadmap is NVIDIA flexing that they're not just leading the AI chip race—they're lapping the competition.
While competitors scramble to match current-gen performance, NVIDIA's already designed the next THREE generations.
The confidence to publish a roadmap this far out shows they're not worried about competition catching up.
What it means:
NVIDIA isn't just selling chips. Nah. They're selling the future of AI compute. This roadmap is basically telling customers "we've got you covered for the next half-decade."
Companies planning their AI strategies now need to factor in these future capabilities. What seems impossible today will be standard in 36 months.
3. Zapier and Microsoft Jump on the MCP Train 🚂
Both Microsoft and Zapier announced support for the Model Context Protocol (MCP), and that's a BIGGER deal than it sounds.
This protocol enables seamless integration between apps and AI assistants without clunky plugins or custom code.
Your favorite AI can now interact directly with thousands of apps through these platforms.
MCP is becoming the universal translator between AI models and the software you use every day.
With these major players on board, MCP is rapidly becoming the industry standard for AI-to-app communication.
The days of building custom integrations for every AI model are dwindling.
What it means:
MCP’s fast adoption could squash the custom AI integration market. Anyone selling proprietary connectors between AI and apps is about to have a very bad year.
The companies that control these integration layers will have enormous influence over the AI ecosystem. Microsoft and Zapier just claimed their territory.
By 2026, we'll take for granted that any AI can seamlessly control any app. MCP is making that possible now.
4. Pennsylvania Saves 8 Hours Per Week Using ChatGPT in Government 🏛️
Pennsylvania's pilot program integrating ChatGPT into government services just reported mind-blowing results.
State employees saved an AVERAGE of 8 hours per week.
That's a full workday reclaimed from bureaucratic busywork.
The program focused on automating repetitive tasks and document processing that previously required manual handling.
This real-world test proves AI can dramatically improve government efficiency without the hype.
For once, taxpayer dollars are being saved by technology instead of wasted on it. lolz.
What it means:
This destroys the argument that AI isn't ready for government use. An 8-hour weekly productivity boost is impossible to ignore, even for the most technology-resistant agencies.
But real talk — if you’re only saving 8 hours a week by using ChatGPT, we thinks ya need some training.
5. Adobe and Microsoft Team Up to Bring AI Agents to Office Apps 🤝
Adobe just introduced Marketing Agent and Express Agent AI tools that integrate directly into Microsoft 365 apps.
These aren't just add-ons—they're fully integrated AI assistants that understand both your content and Adobe's creative tools.
Create professional designs, marketing materials, and presentations without leaving your familiar Office environment.
This partnership merges Adobe's creative AI with Microsoft's productivity ecosystem in a seamless experience.
The days of switching between apps to create professional content are winding down.
These agents understand both the creative and business contexts of your work.
What it means:
Adobe is playing defense AND offense. They're ensuring their creative tools remain essential even as AI threatens to commoditize design.
Microsoft continues its strategy of being the AI integration layer rather than just building models. This is smarter than trying to compete with OpenAI and Anthropic directly.
By 2026, standalone creative apps will feel archaic. Everything will be accessible through AI agents embedded in your everyday tools.
6. OpenAI Makes Sora Unlimited for Paid Users 🎬
Their video-generating tool can now be used without restrictions if you're a paying customer.
This is a HUGE shift from their previous limited-access approach.
OpenAI is clearly confident in both Sora's capabilities and their infrastructure's ability to handle the load.
Video creation is about to be democratized in ways we can barely imagine.
What it means:
The video production industry is facing its "ChatGPT moment." What happened to writers in 2020-2022 is about to happen to videographers and editors.
OpenAI is accelerating their monetization strategy. The previous limited access was about testing; this move is about revenue.
7. Apple Plans to Put Cameras in Your AirPods and Watch 👁️
Apple is reportedly developing cameras for future Apple Watch and AirPods models with integrated AI functionality.
Yes, you read that right. CAMERAS. In your EARBUDS. Yikes.
These wouldn't just be for taking awkward ear-height photos. They're designed to feed visual data to on-device AI systems.
Imagine your AirPods recognizing people you're talking to or your Watch identifying objects you're interacting with.
This represents Apple's unique approach to AI: highly personal, on-device, and integrated into products you already wear.
The privacy implications are both fascinating and terrifying.
What it means:
Apple is betting that the future of AI is personal, not just powerful. They're creating an ecosystem where AI understands your world from YOUR perspective.
But is this too much?
Probably.
This screams “yeah, sounds cool but let’s maybe have a little bit of privacy?”
We get the move by Apple, though, as Big Tech is racing to get as much data as possible to train future world models on how we interact with the real world.
Apple’s got the hardware advantage and is in like last place in the AI race, so they’re hoping jamming cameras everywhere can help them play Chess while they lose in Checkers.
8. Federal Government Launches AI Tool While Workers Worry 🏢
The General Services Administration (GSA) launched a generative AI chatbot to improve efficiency as concerns rise over worker layoffs and surveillance.
Talk about awkward timing.
The new tool is designed to streamline government operations and reduce busywork.
Meanwhile, federal workers are increasingly concerned about job security and AI-powered monitoring.
This tension between efficiency and job security is playing out across the entire government workforce.
The tool promises to eliminate repetitive tasks, but many workers fear it's eliminating their positions.
What it means:
Government AI adoption will create a massive workforce transition crisis by 2026. Agencies need transition plans NOW, not after the jobs disappear.
The public sector faces unique challenges with AI adoption—unions, regulations, and political considerations make this more complex than private industry.
9. Google Gemini Adds Canvas and Audio Overviews 🎨
Google rolled out two major new AI features in Gemini: Canvas for collaboration and audio overviews for document summaries.
Canvas enables real-time collaborative work with Gemini, similar to Google Docs but with an AI participant.
Multiple team members can now work with Gemini simultaneously on the same project.
The audio overviews feature transforms documents into podcast-style summaries that you can listen to while multitasking.
These features show Google focusing on practical workplace applications rather than just raw AI capabilities.
They're building Gemini to fit into existing workflows rather than creating entirely new ones.
What it means:
Google is finally finding its AI product-market fit in their AI suite. After falling behind in the pure capabilities race in 2022 and 2023, they're focusing on integration with their ecosystem.
The audio summaries feature will change how we consume information and the Canvas feature closes the gap with OpenAI’s impressive Canvas offering.
10. OpenAI Unveils New Voice Models That Make Whisper Look Ancient 🎙️
OpenAI announced three new text-to-speech models that are absolutely revolutionizing audio AI.
The lineup includes GPT-4o transcribe, GPT-4o mini transcribe, and GPT-4o mini TTS (text-to-speech).
These models deliver dramatically improved transcription accuracy and voice generation capabilities.
Real-time streaming speech-to-text makes conversations with AI feel natural instead of awkward.
Want your AI assistant to sound excited? Sad? Australian? A cereal-crazed kid?
Just tell it. Customize accents, pitch, tone, and emotional expressions through simple text prompts.
These models outperform OpenAI's previous Whisper model by a massive margin.
What it means:
OpenAI is slowly becoming the do-everything AI company.
The text-to-speech space was ElevenLabs to dominate. Now, they have company with OpenAI’s new offering.
Oh, you can check out OpenAI.Fm for a cool no-code free preview of the new tech.
11. Claude Finally Discovers the Internet Exists 🌐
Anthropic announced a major update enabling Claude to FINALLY! search and process information from the internet in real time.
Welcome to 2023, Claude!
#sorrynotsorry
The new web search feature is available now for paid users in the US.
Instead of relying solely on outdated training data, Claude can now access and synthesize current information just like its competitors have been doing for YEARS.
Anthropic is emphasizing direct source citations, a move designed to address concerns about misinformation and AI hallucinations.
What it means:
Anthropic's front-end gatekeeping strategy will backfire.
Keeping their best models locked away from real-time info while competitors gave public access is a bad move.
This is a desperate catch-up move, not innovation.
Reply