😺 🎙️ PODCAST: How Google's head of AI Studio builds apps in under a minute

Welcome, humans.

Logan Kilpatrick from Google DeepMind just revealed something wild on our podcast.

He says you can take screenshots of websites, drops them into AI Studio, type “clone this,” and gets a working version in seconds.

In fact, that's how Google's own AI team prototypes their own products (Logan literally takes a screenshot of AI Studio itself, spins up a clone, then iterates on it). 

This is something you could do, too. Like, right now! That’s because AI Studio is completely free and already available to you to experiment with.

In our latest podcast episode, Logan (who leads AI Studio and Gemini API at Google DeepMind) pulled back the curtain on Google's entire AI ecosystem—and showed us exactly how to use it:

The actionable gold from this episode:

Getting started with AI Studio:

• (3:31) How AI Studio has evolved: It started as a playground tool, but… “we've moved on beyond that... we released this build tab where you can go in and actually have the model build you an entire app.”

• (4:42) Logan's personal AI Studio workflow: “Prototyping of AI experiences is is how I spend most of my time. It's like, ah, wouldn't it be cool if you could do this thing with AI in a product?…I can just put in a really simple prompt and it just like is connected to my Gemini API keys and it's connected to Gemini and all of our other models by default out of the box…It's just like baked in.”

• (46:03) Logan's go-to first project: “Clone your favorite website. I take a screenshot of AI Studio, put it back in AI Studio, and I say clone this.”

• (16:59) AI Studio’s sweet spot for your workflow: “We're not trying to solve every problem... with all the vibe coding stuff we're doing and with build mode, we’ll get you a working prototype and then get you out into a full-fledged professional developer product.”

Real world applications of AI Studio you can try today:

• (21:43) Text-to-speech applications: “There's been a huge amount of demand... for very native human sounding text-to-speech video generation.”

• (22:18) Make your own AI podcast in your voice: “You could make your own AI podcast in your voice... there's actually not a big gap between being able to do that and what Notebook LM is doing.”

• (26:43) AI tutor for learning: “Having an AI tutor that could help me... would have been so valuable and not spend my time getting yelled at on Stack Overflow.”

• (30:50) A surprising Lawyer adoption stat: “99% of lawyers had tried Gemini, Claude, or ChatGPT.” Why Gemini? “…they just want long context to bring more documents.”

• (36:04) Build a food tracking that actually works: “Snap a picture and have the model deeply understand... using bounding boxes and size proportional understanding.”

• (42:04) Medical applications with MedGemma: “I talk to customers all the time who are blown away with MedGemma... one of the highest traction query use cases in ChatGPT was this medical use case.”

• AI Studio / Gemini’s most underrated features? 

  • Deep Research (47:04): “The model will go visit between tens and hundreds or even thousands of different web pages... it's a mind-blowing experience.” (P.S: in case you forgot, Google invented Deep Research lol).

  • The Live API (20:54): “All the context you need to solve the problem is visible on my screen... the AI system doesn't need access to other tools.” (and looks like OpenAI is competing on this vector now too).

Developer and technical insights:

• (5:45): Google Cloud integration: “AI Studio is built on all the Google cloud infrastructure... deploy uses Cloud Run behind the scenes.”

• (17:31): Large codebase challenge: “There's lots of context buried in different places... it's really difficult to orchestrate this together.”

• (18:08) On context engineering: “Not based on the user prompt... it's trying to do optimization on getting the relevant data to actually help answer.”

• (39:28) Firebase Studio as another solution: “Firebase Studio... you get the batteries included with compute and storage... should be able to get it to do image understanding with Gemini through Firebase, get a database, all that stuff.”

• (40:20): Google’s open source strategy and why Gemma (their small open models) exists: “It's us meeting developers and customers where they are.”

Where all this goes from here:

• (8:26) On Genie 3 and world models: “Fully functional world models that have temporal consistency where you can make edits to the world… it's not like there's a file somewhere being referenced... it is actually on the fly generating the actual ecosystem.”

• (13:03) Logan’s AGI reality check: “The models can't follow basic instructions of chess... they want to make all these illegal moves…I could probably teach a 12-year-old how to play chess in 30 minutes and they'd be able to pick it up.”

• (43:35) What’s Logan most bullish about with AI? Vibe coding: “something that was reserved for a small group of people is now able to be done by a much larger group…The amount of software in the world is going to be a million x what it is right now in 10 years.”

The most exciting moment for us personally? When Logan admitted he's using a product he knows will be obsolete soon because the future version is so much better. We can’t wait to try out the new version of AI Studio when it’s ready!

Watch now on YouTube | Spotify | Apple Podcasts

Resources to dive deeper:

• Try Google AI Studio now (seriously, it's free).

• Google's new Nano Banana image model (here’s our article on it).

• Try Nano Banana yourself in AI Studio.

• Google’s Genie 3 world model (and the other companies building these).

• Logan's convo with DeepMind CEO Demis Hassabis on the future of Genie.

Other Google tools to know:

• Google Gemini.

• Gemini CLI.

• Google Labs (for experimental AI tools).

• Kaggle Game Arena (watch AI compete!).

• Google's Jules (AI coding agent).

• Google Colab (used mostly for training AI models).

They got a lot going on!!

Stay curious, The Neuron Team

P.S. Logan's spiciest take (23:07): “The thing people were most compelled by [in Notebook LM] was the humanness of the scripts... There's no secret sauce... It's literally prompting.”

The AI podcast phenomenon that broke the internet? Just good prompt engineering. Which means you could possibly build the next viral AI tool with what you learn in this episode. Have fun!

P.P.S: Love the newsletter, but don’t want to get these podcast announcements? Adjust your preferences here.