Noah on AI: Issue 18

Let’s get started—plenty to cover this week.

Pretty quiet week in AI and Google wins it with an above average showing. New RAG (Retrieval Augmented Generation) tools, health partnerships, and deals being finalized. Here’s what Google accomplished:

Strong showing for Google. Every time they win I think about how behind they were just a few years ago despite their clear data advantage. Very impressed with their AI growth over time.

Who’s moving, who’s stalling, and who’s climbing: Ordered by production this week.

Interview Outline: Sam Altman discusses the accelerating path toward Artificial General and Superintelligence, the promise of abundant energy and intelligence to reshape society, and why he believes now is the best time in history to start a company. He reflects on OpenAI’s founding, the power of conviction and adaptability, and how startups can move faster than incumbents in an era where “the world is still sleeping on AI.” Altman also shares his optimism for a future defined by innovation, abundance, and human progress.

About the Interviewee: Sam Altman is the CEO and co-founder of OpenAI, the company behind ChatGPT, DALL·E, and GPT models. Before OpenAI, he co-founded the location-based app Loopt at age 19 and later became President of Y Combinator, where he helped launch thousands of startups including Airbnb, Stripe, and Coinbase. Altman is widely regarded as one of the most influential voices in AI and innovation, known for combining bold conviction with pragmatic optimism.

Interesting Quote: “Someday we’ll look back on this moment as the beginning of an era when we started using intelligence itself as a raw material.”

My Thoughts: I always try and listen to at least a bit of a larger interview that Sam Altman is doing. I’m intrigued by the pathway towards artificial general and super intelligence, and believe Altman and OpenAI will be the first or second ones there. Abundance in intelligence and energy leads towards a future that looks very different from today. That’s essentially what all of these AI companies are chasing (outside of $ for some).

Difficulty: Mid-level

There’s all types of focus areas for AI agents, one of the more applicable ones in today’s enterprise world are voice agents. Instead of typing, you can talk to an AI that listens, reasons, remembers, and acts all in real time.

Every AI agent is built around three main components:

🧠 Brain / Interface: the LLM that handles reasoning and conversation.

📚 Memory / Knowledge Base: the information source that stores uploaded files, URLs, or notes for reference.

🛠️ Tools / Connectors: the actions layer that lets your agent schedule meetings, send emails, or do legitimate tasks.

ElevenLabs adds a fourth layer → Speech. It lets you create or clone realistic voices that give your AI a unique tone and personality. Paired with models like GPT or Gemini, your voice agent can hold human-like conversations and take action based on what it hears.

This is what it looks like internally:

Imagine you’re unavailable and a new client rings your business line. Your voice AI agent picks up instantly, greets them, asks about their needs, and checks your calendar availability to book a meeting slot, all without you having to step in.

Voice agents can handle things like customer service, lead qualification, and scheduling tasks. The nice part about AI agents are they can be given enough of a knowledge base to understand how your company operates. They work best when structured for specific roles and built to hand-off tasks to other agents within a system.

Want to build your own? Check out the video guide below:

Adam: Design 3D models with a simple text prompt instead of endless clicks.

The Problem: Traditional Computer-Aided Design (CAD) software is notoriously old, clunky, and complicated. Engineers and designers waste hours on slow, repetitive tasks like fixing geometry, optimizing feature trees, or applying the same changes across multiple parts. The steep learning curve also prevents most people from ever creating their own 3D models.

The Solution: Adam is an AI Agent and copilot that uses natural language to perform complex CAD tasks for you. Instead of manually clicking through menus, you can simply tell the software what to do ("make this hole 5mm wider," or "optimize the feature tree"). It started as a popular text-to-3D generator and is now moving into professional workflows, aiming to automate the most time-consuming work for mechanical engineers in platforms like Onshape.

The Backstory: Founded by Zach Dive and Aaron Li, Adam was a hit in Y Combinator's Winter 2025 batch, quickly going viral with their consumer-friendly text-to-3D app. The company recently raised a $4.1 million Seed round led by TQ Ventures. Their goal is to take the simplicity of AI-native design and apply it to the precision and complexity required by enterprise engineering.

My Thoughts: This is cool stuff. As I pointed out a few weeks ago, the advances we’re seeing in AI are going to lead to a lot more “vibe-coding / vibe-Xing” opportunities. The ability to talk to a computer in regular semantics and have it generate code is the start. Doing the same thing for robotics, design, 3D printing, etc. is where I see a lot of this going and it’s something Adam is taking on.