Sam Nadler (00:00):
Oh, nice party mode. Yeah. So we'll wait for the beat drop because I made this song custom as well. Sounds—yeah.
Jordan Metzner (00:06):
Sam, you're probably familiar with what's called an applicant tracking system, or ATS for short. Basically, it's a platform to know where your candidates are for every step of the process.
Sam Nadler (00:15):
It seems like Envato themselves have gotten into the AI space.
Jordan Metzner (00:18):
On Friday, The Information announced that Meta is shaking up its AI org again. We haven't seen a new model from them in a while. Obviously, Elon’s been spending hardcore on getting Grok out to users.
Sam Nadler (00:28):
This Week—breaking it down. We show you how fresh AI can be a clever tweak you like. What did you build this week?
Sam Nadler (00:32):
Hey everyone and welcome to another episode of Built This Week, 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, co-founder of Ryz Labs, and I'm joined by my friend, co-host, and business partner Jordan Metzner. What's up Jordan?
Jordan Metzner (00:48):
Hey Sam, I'm super excited to be back. Another new episode, another crazy week in AI. Lots of leveling up this week.
Sam Nadler (00:56):
Before we dive in, please remember to like and subscribe. Whether you're on Spotify, Apple Podcasts, or YouTube—I'm personally a YouTube fan—sign up and hear us every week.
Sam Nadler (01:06):
The docket today: I built a tool, a front-end prototype I’ll hopefully pass to our dev team soon. It’s pretty intense but I built it fast—my first time using Cloud Code. We'll dive into that. We’ll also demo a tool we use all the time at Ryz Labs, and then wrap with top AI news from the week.
Jordan Metzner (01:25):
Let’s just jump in. You ready to share your screen?
Sam Nadler (01:28):
Let me give a little context first. At Ryz Labs, one of our core businesses is technical staffing. If you’ve been at a large organization hiring, or if you’ve been a hiring manager, you probably know what an ATS is—applicant tracking system.
Sam Nadler (01:43):
So I thought—why not build one? An ATS customized for our exact use case.
Sam Nadler (01:47):
I started with a generic PRD in OpenAI and threw it into Cloud Code. A few bumps early on, but after some back-and-forth with Cloud Code, I got the first version done. Then I refined it: second PRD, detailed use case, listed the key features I liked from our current ATS, made sure those were built in.
Sam Nadler (02:08):
What’s amazing about Cloud Code is I could spin up multiple agents working on different features in parallel. One agent worked on the jobs page, another on the candidate page, another on something else. I topped out at around 30 agents running in parallel.
Sam Nadler (02:26):
Let me run you through it. There’s a jobs page with roles posted. A candidates page—this is all fake data—and a pipeline board where you can move candidates across stages. Looks like a Kanban board with draggable cards.
Jordan Metzner (02:39):
Can you describe what it looks like for listeners?
Sam Nadler (02:41):
It’s beautifully designed. Cards show candidate names, status, tags. You can move them across stages. You can edit the stages too. There’s a page showing all candidates and where they are. You can filter them—say, all sourced candidates in the last hour.
Sam Nadler (02:59):
There's an Offers page where you can create offer templates. A Feedback page where you customize interview forms depending on the role—technical or not. There's a page for analytics. A Tasks section for your recruiting team—schedule interviews, send offers, etc.
Sam Nadler (03:16):
And, of course, I added light mode, dark mode... and party mode.
Jordan Metzner (03:20):
Nice. Party mode.
Sam Nadler (03:22):
We’ll wait for the beat drop. I made the song custom too.
Jordan Metzner (03:25):
Incredible.
Sam Nadler (03:26):
Anyway, my next step is to polish each page, test every button, and pass it to our engineers. It’s a big system, so it’s not a quick build, but maybe a couple months?
Jordan Metzner (03:38):
Maybe even faster. But before handing it off—how valuable is this as a prototype to test with the recruiting team?
Sam Nadler (03:45):
Super valuable. With quick iteration cycles, I could get feedback from recruiters during a call and implement it live.
Jordan Metzner (03:50):
And with Cloud Code, we could design the backend too—database structure, edge functions—so when we hand it to engineering, they’re really just implementing.
Sam Nadler (04:00):
Exactly. I haven’t used Cloud Code for backend yet, but I assume it’s just as good. The more context we give—schema, functions, PRDs—the faster they can implement.
Jordan Metzner (04:12):
You could even record feedback sessions and run the audio through ChatGPT. Screenshot each page, collect feedback, generate tasks.
Sam Nadler (04:18):
Yeah, great point. On AI features—we’ve already built a resume scoring tool called RyzScore. We could bake that in. We could also layer on automated emails, job matching. Lots of candidates don’t get placed right away but might be great for future roles.
Jordan Metzner (04:35):
Exactly. Match them quickly. And with these short iteration cycles, you could iterate and re-prompt Cloud Code with recruiter feedback minutes later.
Sam Nadler (04:45):
And do it again. And again. And again.
Jordan Metzner (04:47):
Exactly. Once you're aligned with recruiters, then start on the backend.
Sam Nadler (04:50):
Right. Don’t build backend until core features are locked in.
Jordan Metzner (04:53):
Sam, great job. Seeing your evolution—from Bolt to using 30+ agents in Cloud Code—in just 8 weeks is incredible.
Sam Nadler (05:00):
Thanks. It depends on the use case. Bolt is great for quick builds, landing pages. But for something this big, Cloud Code was ideal.
Jordan Metzner (05:08):
And now you can choose tools based on what you're building. Bolt, Replit, Cloud Code—they all have different strengths.
Sam Nadler (05:14):
Absolutely. And hopefully I’m proof—if I can do this, anyone can.
Sam Nadler (05:17):
With that, let’s move on to our tool of the week.
Jordan Metzner (05:20):
Envato. I've used their products for 10–15 years. We use Envato Elements—a library for stock video, audio, and graphics. Historically, it’s been great for music and templates. But now they’re getting into AI.
Sam Nadler (05:34):
Let’s test their AI image tool. Create an Instagram ad for Built This Week, promote new episodes every Friday. Enhance prompt? Sure.
Jordan Metzner (05:43):
It looks like it’s running across multiple models. But honestly, none of these look great. Text is jumbled. Faces are wrong.
Sam Nadler (05:52):
Let’s try our own internal image gen tool instead.
Jordan Metzner (05:55):
Same prompt, but now with the Google model. This is way better. The text is readable, the layout is clean.
Sam Nadler (06:00):
Let’s run the prompt across several models. This one’s from Ideogram, this one from Google, Flux, ByteDance… Yeah, much better results.
Jordan Metzner (06:09):
We’ll still use Envato for stock footage when we need it, but for generative stuff, our internal tools are ahead.
Sam Nadler (06:15):
It’s a hard problem. Still early days for generative image tools.
Jordan Metzner (06:18):
Let’s get into the news. Meta’s shaking up its AI org again. What’s going on?
Sam Nadler (06:22):
They haven’t released a new model in a while. Zuckerberg’s reorganizing teams to move faster. Heard there were cultural tensions—some engineers making huge comp packages, others not as much.
Jordan Metzner (06:34):
Yeah, that’s tough. In sports, it’s clear who’s better. In software, it’s murkier.
Sam Nadler (06:38):
Exactly. Lines blur. Hierarchies get political. Some people feel undervalued.
Jordan Metzner (06:42):
What else is happening?
Sam Nadler (06:44):
Elon’s pushing Grok hard. But public share links from Grok were indexed by Google, revealing some user data. Same thing happened with ChatGPT recently.
Jordan Metzner (06:51):
Nothing gets by Google. Their indexers are sharp.
Sam Nadler (06:54):
Have you used Grok?
Jordan Metzner (06:55):
A little. It’s okay. Not better than Claude for coding or GPT-5 overall.
Sam Nadler (06:59):
Well, great episode, Jordan. Thanks for walking through everything.
Jordan Metzner (07:02):
Thanks for listening. Like and subscribe to Built This Week. We’re here every Friday.
Sam Nadler (07:06):
YouTube, Apple, Spotify—wherever you get your podcasts. Thanks everyone!
.png)
