Sam Nadler (00:00):
Happy Halloween! Let’s take a picture of you. You’ll be able to pick a pre-designed costume, or type in your own costume.
Sam Nadler (00:14):
Hey everyone, and welcome to Built This Week, Episode 18 — the podcast where we share what we’re building, how we’re building it, and what it means for the world of AI and startups. I’m Sam Nadler, co-founder here at Ryz Labs, and each week I’m joined by my friend, business partner, and co-host Jordan Metzner.
Jordan Metzner (00:39):
Yo Sam, how’s it going? Episode 18 — another crazy week in AI! Massive announcements from the hyperscalers, and some really fun tools to share this week.
Sam Nadler (01:07):
Yeah, this one’s even seasonal. Before we get into the agenda, quick reminder to like and subscribe — we drop new episodes every Friday on YouTube, Spotify, and Apple Podcasts. This week we’re highlighting a fun internal tool we built called the Trick-or-Treat Ryz Tool. It’s part cultural, part seasonal — uses Nano Banana and Veo 3.1.
We’ll also talk about Publisher, one of our favorite internal tools to distribute and post content. And finally, we’ll cover this week’s AI news: OpenAI’s latest products and a few funny stories.
Jordan Metzner (02:19):
Before we jump in — if you could share the podcast with a friend or two, that’s the best way to help us grow. We’re almost at 14,000 subscribers on YouTube and nearing our 20th episode!
Sam Nadler (02:57):
Perfect. Let’s start with Trick-or-Treat. You built this tool for teams — walk us through it.
Jordan Metzner (03:28):
Sure. The idea came from Paige Bailey — she’s the dev lead at Google DeepMind. She posted a demo using a webcam with Nano Banana and Veo 3.1. I thought it was a fun idea, so we built our own version. It’s not the most creative name — “Trick-or-Treat Ryz Costume Builder” — but it works!
Basically, you take a webcam photo (on desktop or phone), choose or type a costume, and the app uses Veo 3.1 to generate a “Happy Halloween” video of you in costume.
Sam Nadler (05:10):
All right — should we try pirate, superhero, ninja, or custom?
Jordan Metzner (05:12):
Let’s go ninja. It takes about 20–40 seconds — and… there I am, a ninja! Not quite my face, but my body and movement. Then it makes a short Halloween video.
Sam Nadler (06:30):
That’s hilarious. Let’s try another one.
Jordan Metzner (07:11):
Okay, how about an 1800s cowboy? I’ll pose… boom, done. It’s fast — background removed, pose kept. Then it generates a cowboy video.
People in the office have done vampires, wizards, everything. It’s fun seeing how far edge video generation has come.
Sam Nadler (08:14):
It’s been a blast. One teammate even got a real Halloween costume idea from it. But talk about how you built this — I know you used Replit and maybe the new planning feature?
Jordan Metzner (09:45):
Yeah. I built it in Replit using the Nano Banana and Veo 3.1 APIs, served through Replicate.com. It’s simple — webcam captures image → sends to Nano Banana → returns new image → sends that to Veo 3.1 → generates a short themed video.
It’s not perfect — the characters sometimes mumble fake words — but it reminds me of early image generation where text was gibberish. It’ll improve fast, especially with Nvidia’s upcoming video chips.
Sam Nadler (11:29):
Cool. Let’s move to Publisher, the social media syndication tool we use.
Jordan Metzner (11:56):
Publisher lets us post systematically across multiple channels — LinkedIn, YouTube, Shorts, X, TikTok — all from one dashboard. You can queue posts, use AI to write captions, add hashtags, schedule, and track analytics.
Without it, our social team would spend hours manually posting. With Publisher, we can scale easily.
Sam Nadler (14:00):
Exactly. It supports all major platforms, team access, previews, scheduling — and we use it constantly across Ryz Labs and our sub-brands. It’s been a massive time-saver.
Jordan Metzner (16:18):
Yeah, we used Buffer before, and Canva also has a scheduling tool. These platforms are a must for startups that want to scale content without big teams.
Sam Nadler (16:59):
Also, it saved us during a teammate’s European vacation — everything kept posting seamlessly.
Jordan Metzner (17:47):
Exactly. The brands never stop posting!
Okay — news time. Our first story: Kohler, the toilet brand, has unveiled a new camera for your toilet.
Sam Nadler (18:14):
Yes, a toilet camera. Supposedly for health analysis. Interesting idea — maybe too much optimization?
Jordan Metzner (18:38):
Yeah, it sounds gross, but analyzing waste can reveal a lot. This might be the start of AI entering the bathroom — smart toilets, sensors, full health data integration. Weird but potentially useful for medical monitoring.
Sam Nadler (19:47):
Still, a stinky story. Let’s move on.
Our next headline: OpenAI launches Atlas, its new web browser.
Jordan Metzner (20:05):
Right — Atlas dropped yesterday. It makes sense strategically — OpenAI building its own browser. Google even took a small stock hit.
Sam Nadler (21:00):
I’ve used it for about 20 minutes. It’s super fast — lighter than Chrome — and tightly integrated with ChatGPT. Searches bring you directly into a chat instead of a Google results page, which is a huge shift.
You can see how shopping or travel could move entirely inside the chat experience.
Jordan Metzner (22:53):
Exactly. It’s like having an assistant browsing for you — summarizing pages, highlighting the best parts, analyzing data. Early days, but you can see the direction: talk to the web instead of browsing it.
Sam Nadler (23:29):
Yeah. I tested it by asking Atlas to recommend a family passport holder. It asked a few questions, offered two or three Amazon options, and explained the pros and cons. Felt like shopping via conversation.
Jordan Metzner (24:35):
That’s the key — memory. Once it remembers preferences, it can handle recurring tasks like booking Airbnbs, ordering groceries, or planning travel automatically. That’s when it becomes a true agent.
Sam Nadler (26:19):
Agreed. Okay, last story: Google’s new AI Studio and its “Build” feature.
Jordan Metzner (26:43):
Yeah — they launched it yesterday. It’s basically a browser-based AI app builder powered by Gemini 2.5 Pro. It’s like Lovable, Bolt, or Vercel’s AI tools — but with Google’s polish.
You can start from templates (image-to-video, chatbot, maps, transcription, etc.) and deploy instantly.
I tried building a weather-and-time app for LA and Nashville. It generated, styled, and deployed it in seconds. You can push to GitHub, preview on mobile, or export the code.
Honestly, huge step forward for Google. Combine this with Atlas, and it’s clear: AI interfaces are converging. Soon, we’ll just describe what we want — and it’ll be built.
Sam Nadler (30:01):
Awesome. Great episode, Jordan. Everyone, like and subscribe — and we’ll see you next week.
Jordan Metzner (30:22):
Thanks everyone. Happy Halloween!
