Is NVIDIA the New Kingmaker of AI?

In partnership with

Hi everyone,

This is Lore, the newsletter that helps you thrive in the age of AI.  

I hope you had a great New Year and are ready for 2025. Things are about to get wild.

Thank you for being a part of 23k+ techno-optimists and growing. 🙏

P.S. Watch the latest episode of The Next Wave where we made our predictions for 2025!

Want to get the most out of ChatGPT?

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide with actionable prompts and tips to make your work more productive.

1) Things Are Getting Better

2) Is NVIDIA the New Kingmaker of AI?

At CES 2025, NVIDIA unveiled moves that could redefine the AI landscape. The headliner? Project DIGITS—a $3,000 personal AI supercomputer built around their cutting-edge GB10 Grace Blackwell Superchip. This compact powerhouse delivers a full petaflop of AI performance, boasting 128GB of unified memory, and comfortably runs models with up to 200 billion parameters from a standard wall socket.

The big question: Will NVIDIA’s strategy of putting AI supercomputers on every desk challenge the current cloud-centric status quo?

I recently hashed this out with my co-host Matt Wolfe on The Next Wave. He’s amped up about a future in which private, locally hosted AI models become standard—ensuring users maintain strict control over data while still reaping advanced capabilities. It’s a compelling vision for those who want the power of large language models without sending sensitive prompts into the wild.

Still, I’m skeptical this spells doom for the giant cloud-based players like OpenAI and Anthropic. Training the largest, most sophisticated models often demands more computing resources than most individuals—even well-funded startups—can realistically handle. My bet is we’ll see more hybrid solutions: let the cloud handle massive training tasks, while personal AI hardware takes care of specialized fine-tuning and day-to-day inference.

Regardless of how it plays out, there’s no denying that NVIDIA is extending its reach across every layer of AI infrastructure, from consumer gear to enterprise-scale solutions. Beyond DIGITS, NVIDIA’s Cosmos platform offers generative ‘world foundation models’ specifically tailored to robotics and autonomous vehicles. These models create physics-aware video from minimal inputs, helping developers simulate complex real-world scenarios and minimize testing risks.

What do you think? Will personal supercomputers usher in a wave of locally trained, privacy-friendly AI, or does the cloud remain king for the foreseeable future?

AD

Your daily AI dose

Mindstream is your one-stop shop for all things AI.

How good are we? Well, we become only the second ever newsletter (after the Hustle) to be acquired by HubSpot. Our small team of writers works hard to put out the most enjoyable and informative newsletter on AI around.

It’s completely free, and you’ll get a bunch of free AI resources when you subscribe.

Click here to subscribe

3) AI Predictions For 2025

In this week’s episode of The Next Wave, Matt Wolfe and I made our predictions for 2025. We talked about Agents, AGI, ASI, and a lot more. One thing is for sure: it will be a crazy year. Check out the episode!

Also, please subscribe to The Next Wave on Apple, Spotify, and YouTube. 

4) RepoPrompt: My Secret Weapon for Supercharging AI

Over the past week, I stumbled on a tool that’s completely transformed how I work with AI—one that’s easily tripled my productivity. Today, I want to introduce you to RepoPrompt (no affiliation), show you my own use cases, and spark ideas on how you might leverage it, too.

What Is RepoPrompt and Why Is It So Game-Changing?
RepoPrompt (Mac-only for now) allows you to create extra-long prompts for AI models. If you’re using a model like ChatGPT o1 pro—which can be incredibly powerful once you feed it the right context—you need a way to supply enormous chunks of code and instructions. Without that context, you wait forever for a bland response.

How I’m Using RepoPrompt
I’m building a game and using ChatGPT o1 pro for almost all the coding. Previously, I had a cumbersome setup:

• A script to compile all my code across many different files into one massive text file.

• Multiple “instruction” text files telling ChatGPT o1 pro how to respond (like “Please return diffs or code blocks,” or “You’re an expert engineer, keep it concise.”).

• Frequent toggling between files, copying and pasting, and re-running scripts to keep everything updated.

Since switching to RepoPrompt, I can simply point it at my project folder, designate which files to include, and save different instruction sets. RepoPrompt automatically handles the context, so ChatGPT o1 pro knows exactly what my project looks like and which style of response I expect.

The Biggest Win
RepoPrompt can instruct ChatGPT o1 pro to return an XML patch, which I can merge back into my code with a couple of clicks. Gone are the days of scanning ChatGPT o1 pro’s suggestions line-by-line and manually integrating them. Now, I speak into Flow (an AI voice transcriber) to generate the prompt, paste the output into RepoPrompt, and watch it work its magic—over 50,000 tokens of context, if necessary.

Why This Matters

• Speed: I no longer spend half my time copy-pasting code.

• Accuracy: With a fully loaded context, ChatGPT o1 pro’s suggestions are sharper—and faster to implement.

• Scale: Large projects that used to baffle certain models now become tractable.

Honestly, I see a future where I can build 95% of an entire project using ChatGPT o1 pro, so long as I keep the instructions crystal clear.

Other Use Cases

• Writing & Editing: Use RepoPrompt to feed ChatGPT o1 pro your style guidelines, writing samples, or references from your favorite authors and creators.

• Social Media Campaigns: Save your best posts or brand guidelines, so the model knows exactly how you want content shaped.

• Research & Data Analysis: Provide relevant documents and context so the AI doesn’t hallucinate or lose track of crucial facts.

Long story short, if you need to feed an AI model loads of context—whether it’s code, text, or any other kind of file—RepoPrompt it’s incredible. Check it out, and you might discover you can push AI much further than you realized.

That’s all for today. Please consider sharing the newsletter with your friends if you think they’d enjoy it. 🙏

Please give me feedback on Twitter. I am still figuring out the best format for this newsletter. Which part did you like? Dislike? Do you have other suggestions? Please let me know—just tweet @NathanLands.

-Nathan