😺 Build AI apps w/ Claude or Google

Welcome, humans.

NEW PODCAST ALERT! Corey and Grant break down the Claude Opus 4 “blackmail” safety test, unpack OpenAI’s efforts at AI transparency, and offer a practical framework to separate genuine AI safety concerns from hype.

You can listen and/or watch on YouTube, Spotify, and Apple Podcasts.

Here’s what you need to know about AI today:

• Claude and Google released dueling AI app building tools.
• NVIDIA is going all in on robots.
• Meta, Microsoft, and Stability AI had new legal action.
• Google released AlphaGenome to predict how DNA mutations affect genes.

Advertise in The Neuron here.

The Day of Dueling AI Builders… Claude's Shareable Apps vs. Google's Free Coder

Ever wanted to build an AI app… using AI? (very meta). Well, it just got A LOT easier.

Yesterday was a day of dueling “vibe-coding” launches; Google released Gemini CLI, a free tool to code with AI by chatting with your terminal, and Claude now lets you build and share AI-powered apps.

Here’s how building apps with Claude works: Y’know how OpenAi has “GPTs”? Claude’s version is Artifacts. Now, your Artifacts can actually run Claude inside them.

Claude writes the code, hosts your app, and lets users access it through a shareable link where their usage counts against their own subscription, not yours. Here’s a demo.

The best part is you can use this no matter what plan you have (free, Pro, or Max).

Here’s an example of some of the templated things you can build with Artifacts:

If you want a more technical breakdown on why this is a big deal and how to best prompt it, read Simon Willison’s instructions.

To turn it on, go to Settings > Profile > Feature Preview and toggle on “Create AI-powered artifacts.”

Now, there’s actually quite a few apps that work like this: Like Lovable, v0, and Bolt.new (to name a few).

…But none of them that are as easy to set up with AI powered by AI.

The closest might be Google’s Firebase Studio, which we’ve written about before, and a new-ish feature we’ve been playing in AI Studio called “Build apps with Gemini” that also lets you build AI-powered apps for free.

What’s really cool about building apps in AI Studio is you can download the code when you get it to a good place, or host them through Google cloud (though that doesn’t seem as easy as doing it with Claude). If you’re interested, this is a great tutorial.

For the coders: Google launched Gemini CLI, which brings AI assistance directly into your terminal, helping you write code, debug issues, and solve problems, and it actually writes and runs the code (code).

It’s basically Google’s answer to the ever more popular Claude Code tool, which lets you code with Claude directly in your computer’s terminal.

But here’s what’s REALLY wild about this: Unlike Claude Code, Gemini CLI has:

1. 1M token context window (roughly equal to ~35K lines of code).
2. 60 model requests per minute.
3. 1,000 requests per day.

…ALL FOR FREE. That's more free stuff than you’d get at a timeshare convention!

Keep in mind, free = you are the product. Google will still collect this data for training purposes, but you CAN opt out.

Our take: Whether you are a newbie with an idea or a dev with deep domain knowledge, just about everybody has the ability to natively build apps now… you’ve gotta wonder, where’s OpenAI’s version? OpenAI has Codex, but where’s its version of Firebase Studio or Artifact apps? Surely they have something cooking in this area…

FROM OUR PARTNERS

As a founder, you’re used to sprinting—toward product-market fit, your next round, or that first enterprise contract.

But with AI accelerating how quickly startups build and ship, security expectations are higher earlier than ever. Getting security and compliance right can unlock growth—or stall it if you wait too long.

That’s why fast-growing startups like LangChain, Writer, and Cursor trusted Vanta to build a scalable foundation from the start.

With deep integrations and automated workflows built for fast-moving teams, Vanta gets you audit-ready fast—and keeps you secure with continuous monitoring as your models, infra, and customers evolve.

Join Vanta's Live Demo on July 10 to see the product in action and ask questions in real time.

Save your spot.

Prompt Tip of the Day

Forget “prompt engineering”, it’s time to focus on “context engineering.” AI experts like Andrej Karpathy are pushing this term to reframe how to work with AI because it better describes the challenge: giving your AI all the context it needs to solve your task.

He says what matters the most when getting AI to help you is assembling the right context (examples, background info, tools, conversation history) to reduce the “fog of war” for your model.

Think of it like prepping for a presentation: too little info and you'll fumble; too much irrelevant stuff and you'll lose focus.

So don’t spend too much time trying to craft the most clever prompt. Focus on giving it thoughtfully curated context that sets your model up to win.

Treats To Try.

*Anything marked with asterisks is sponsored content. Advertise in The Neuron here.

1. *Speak; Flow turns voice into flawless text up to 5× faster. Free 2k words weekly. “Best AI product since ChatGPT”—Superhuman CEO. Download today.
   1. P.S: They just raised $30M! Saw this in the news, had to shout it out.
2. OpenRouter gives you one API to access 400+ AI models, automatically routing to the best option based on your needs (raised $40M).
3. Harvey automates legal work like contract review, document drafting, and legal research using AI trained specifically for law firms (raised $300M).
4. Make Grid gives you a real-time visual map of your entire automation landscape so you can see how all your apps and workflows connect.
5. FlashDocs creates personalized sales presentations in seconds by pulling data from your CRM, call notes, and client background into Google Slides
6. Automaticall wastes scammers' time by forwarding spam calls to an AI that pretends to be a real person and keeps them talking in endless circles.

See our top 51 AI Tools for Business here!

Around the Horn.

• Google DeepMind released AlphaGenome (paper), which helps you predict how specific DNA mutations will affect gene regulation and disease risk.
• NVIDIA sees robots as one of their two largest growth areas—already, its “robotics unit” made $567M for the quarter in May, which is 1% of total revenue.
• Lots of legal action yesterday (after tuesday’s fair use decision):
  • Meta won its own major legal victory after a federal judge ruled again that training AI models on copyrighted books constitutes fair use.
  • Microsoft was sued by a group of authors for its unauthorized use of 200K books to train its Megatron model (also created with NVIDIA).
  • Getty dropped its main copyright infringement claims against Stability AI in its UK-based lawsuit.
  • OpenAI CEO Sam Altman and COO Brad Lightcap had an awkward convo live on stage with NYT podcasters ahead of OpenAI’s next court session with the NYT (which is happening today!)

FROM OUR PARTNERS

Insure Your Cloud Spend with Archera  

AI agents and applications are pushing cloud costs higher by the day. Archera helps keep tabs on your cloud costs in AWS and Azure, but also insures your cloud commitments. That means if you underutilize your commitment, Archera will put money back into your bank account.

Check out how Archera can help you save today.

Thursday Trivia

One is real, and one is AI. Which is which? (vote below!)

A.

B.

Which is AI?

The answer is below, but place your vote to see how your guess compares to everyone else (no cheating now!)

A Cat's Commentary.

Trivia Answer: A is AI (from MidJourney), and B is real.