The Next Wave – AI Tool Better Than OpenClaw? + NVIDIA’S $1T Prediction & AI Image Wars

Welcome to another special episode of The Next Wave Podcast. This episode was packed with big news and fresh ideas about AI, the companies behind it, and how it’s changing our lives. I joined Matt Wolfe as we explored new AI tools, NVIDIA’s huge predictions, the ongoing AI image wars, and where robots and real-world jobs are headed.

What’s the Episode About?

We kicked off with some news: this was our last episode for a little while, as we’re taking a break. But we wanted to share everything we learned at NVIDIA’s GTC conference and cover the topics that made waves this week. We talked about big ideas in AI tech, the new tools emerging, how AI is growing across every part of our lives, and even how robots are beginning to handle tasks that seemed impossible a few years ago.

The episode moved through several main subjects—NVIDIA’s growth, OpenClaw and NemoClaw, brand new AI image models, changes in the job market, and the rise of real-world robots. You’ll get a clear sense of what’s happening and why it matters.

NVIDIA’s GTC Conference and $1 Trillion Prediction

NVIDIA’s GTC conference set the tone for the week. Matt Wolfe described it as the “Super Bowl for AI” and the “Burning Man for AI,” showing how everyone in the field was paying attention. NVIDIA stands at the very center of the AI world, much like the sun in a solar system.

Jensen Huang, NVIDIA’s CEO, shared some jaw-dropping stats. He said that over the past year, NVIDIA sold $500 billion of chips—just in one year! He predicts that chip sales will double, reaching $1 trillion by the end of 2027—possibly even more. What’s wild is that these numbers are tied to actual purchase orders from companies ready to buy as soon as chips hit the market.

NVIDIA’s growth comes from several directions:

• The move from pre-training AI to post-training and inference, which needs tons of compute power.
• Businesses choosing to build their own in-house data centers.
• New types of chips focused on faster inference, thanks to companies like Groq (which NVIDIA basically owns).

Jensen also said that NVIDIA will use half its free cash flow for dividends and buybacks, following the “Apple playbook” to make stock more valuable.

Key Takeaways:

• NVIDIA’s chip sales are projected to hit $1 trillion by 2027.
• Growth is coming from both enterprise and everyday AI tools.
• AI compute needs are shifting towards inference and post-training.
• NVIDIA works with nearly every big company—Google, AWS, Oracle, and more.

OpenClaw, NemoClaw, and Personal AI Agents

OpenClaw got huge attention at the conference. Jensen called it as important to the future as the web browser was to the internet. OpenClaw lets anyone easily set up their own AI agent that acts as a personal assistant. With just one command, it can start helping manage your email, tasks, and more.

NVIDIA introduced NemoClaw, a package that builds on OpenClaw. It adds security, privacy, and better agent handling, plus easy options to install open-source models like NeMoTron 120B. OpenClaw and NemoClaw look set to become tools everyone uses in daily life, powering agents behind Siri, Alexa, and even controlling smart devices at home.

Jensen and NVIDIA believe we’ll all have AI assistants soon—smart enough to control everything from your phone to your toaster.

Key Takeaways:

• OpenClaw makes AI agents simple for anyone to use.
• NemoClaw improves security and privacy for AI agents.
• Personal AI assistants could become as normal as Siri or Alexa.
• Growth in AI will come from widespread agent use, requiring more compute power.

AI Image Wars: Midjourney, Nano Banana, and Microsoft MAI-Image 2

The battle for best AI image generator heated up. We looked at Midjourney V8, but found it still struggles with realism—especially fingers and hands. When we tested hand images, Midjourney gave us some weird results. Nano Banana, a newer tool, handled hands much better. Microsoft’s MAI-Image 2 model also passed the “hand test” and could generate accurate text in images.

These models each have strengths. Midjourney stands out for creative, fantasy, and avant-garde art. Nano Banana is strong with realism and hands. MAI-Image 2 can handle text and looks good for realistic pictures, too. Microsoft’s tool even has an API (though not public yet), and ranks high on leaderboards in blind image tests.

Midjourney’s slow progress may come from its founder refusing to take venture capital or release an API, limiting how other platforms can use it.

Key Takeaways:

• Midjourney V8 is creative, but still struggles with realism and hands.
• Nano Banana passes hand tests and is fast.
• MAI-Image 2 from Microsoft handles hands, text, and realism well.
• Open APIs help new models grow faster.
• The field is rapidly improving, and older tools can fall behind.

AI’s Impact on the Job Market and Digital Exposure

We explored a job market visualizer made by Andrej Karpathy (former OpenAI and Tesla engineer). This grid shows trends for 342 different jobs in the US, coloring them by whether they’re likely to grow or shrink.

Green boxes show growing jobs (like cooks, health care aides, construction workers). Orange and red boxes highlight declining jobs (like office clerks, bookkeepers, customer service reps). When you switch to the “digital AI exposure” layer, the effect of AI becomes clearer—software developers, paralegals, and even top executives could be highly exposed to automation.

Lawyers are a special case: stats show growth, but AI exposure is high. Physical jobs like janitors and construction are less at risk for now.

Key Takeaways:

• Some jobs are shrinking, like bookkeepers, due to automation.
• Real-world data doesn’t always match AI exposure levels yet.
• Software developers have high AI exposure but are still growing.
• Jobs needing hands-on or physical work remain safer from AI.

SaaS Companies and How Agents Change Software

Jensen Huang’s opinion is that SaaS software isn’t going away. Instead, agents (powered by OpenClaw and similar tools) will start interacting with software like HubSpot or Stripe instead of us doing it ourselves. This means workflow software with strong “data moats” (lots of private data and connections) will keep growing.

Simple SaaS tools—just wrappers around APIs—may disappear as agents can build custom tools easily. For example, apps like ThriveCart (simple shopping cart tools) may fade as agents handle tasks through direct API access.

Key Takeaways:

• SaaS incumbents with deep workflows and data moats will survive.
• Simple one-trick SaaS tools could get replaced by custom agent-built solutions.
• Agents may use SaaS tools on our behalf, changing how we work.
• Rapid change is likely as AI tools become easier for everyone.

Anthropic’s AI Survey: What People Want from AI

Anthropic surveyed 81,000 people to learn what they want from AI. Top responses were:

• Professional excellence (offloading routine work)
• Personal transformation
• Better life management (helping with schedules and organization)
• Time freedom and financial independence
• Societal transformation

Most people (81%) said AI has already helped them in at least one way, mainly speeding up productivity. However, AI hasn’t delivered for everyone. Some worry about unreliability or even cognitive atrophy (getting worse at thinking by relying on AI too much).

Emotional support from AI gets a mention, but it’s considered risky for users.

Key Takeaways:

• People mostly want AI to handle boring tasks and help them be more productive.
• Time freedom is valued, but many end up just doing more tasks faster.
• Unreliability is a big concern—especially for work.
• Education around AI use is still needed for many.

DoorDash Tasks and the New Data Economy

DoorDash started offering “tasks” for side hustle money. You can get paid to take pictures at restaurants or stores, film yourself, or record translations. This data trains AI and helps robots understand the physical world.

DoorDash isn’t alone—Uber and other companies offer similar “data gigs.” These jobs aren’t traditional work but help build the next AI wave. Concerns include privacy and a “race to the bottom,” where pay drops as more people join.

Key Takeaways:

• Companies are paying people for real-world data to train AI and robots.
• New types of work focus on data collection, not traditional tasks.
• Privacy and fair pay are important issues to watch.
• Text data is already plenty; now companies need video, images, and audio.

Robots Doing Real Work: Athletic Humanoids

We closed out with a look at athletic humanoid robots. Researchers in China created GalBot—a robot that plays tennis against humans using a learning algorithm. Robotics is improving fast; teams showcase robots dancing, flipping, and now playing sports.

Real-world learning (not just pre-programming) lets robots handle changing environments. Still, robot “brains” haven’t caught up fully with human problem-solving, so some physical jobs remain safe for now.

Key Takeaways:

• Robots can now handle complex, athletic tasks like tennis.
• AI and robotics work together to respond to real-world changes.
• Narrow AI lets robots shine in specific tasks; general-purpose robots aren’t ready yet.