I built an apartment hunter with Claude Code

Why you're getting this: You've interacted with Haven or AdTribe before, and this is my weekly-ish update on my journey to building an AI-native agency. Zero pressure to stick around, just click unsubscribe if you don't want to get it.

Alright, let’s get into it.

My girlfriend and I are looking for a new place in Vancouver. Anyone who's rented here knows good listings disappear within hours. So I built a rental scout in one sitting with Claude Code.

Here's how it works:

Every 20 minutes it checks Craigslist for new listings matching our criteria. When it finds one, it sends it to Telegram with photos, price, neighborhood, and a link. It also checks my email to see if we already applied, so I'm not double-submitting.

So far it's tracked 2,333 listings, sent me 76 that actually matched, and the whole thing costs a few cents a day to run.

I also built a deal detector that flags anything priced significantly below market. The best listings float to the top automatically.

I could go one step further and have it send inquiry emails on my behalf. But then I'd probably end up with a ton of spam from real estate agents haha.

P.S. We already found the house 😍

We recently hired two designers as part of our new offering: building the creative department for our agency partners.

Creatives are everything in paid social right now. And honestly, we've been behind. We have great designers we work with, media buyers who understand creative strategy, but the quality could be better. The output could be higher. Having someone in-house changes that.

So we did it. Hired two exceptional designers, both above market rate.

Now we're building the internal process to justify those salaries and make sure our partners get both a good deal and the best quality output.

One thing I told them early: don't rush. Follow the briefs from media buyers, but spend extra time on the website. Learn the product. We attach marketing research to every brief, so designers should know the product, know the consumer, know what problems they're facing. Then craft the creative that actually resonates.

We're also giving every designer access to a performance dashboard. They'll see how their creatives perform and iterate from there. At the end of the day, we need creatives that outperform the benchmark in each account. Not just beautiful work.

Stay tuned. Within the next month or two we'll be able to offer our partners a much better offer on creative.

AI can see every test we've ever run. Every result, every number, every outcome. The one thing it can't see is why we thought it was worth running in the first place.

That reasoning lives in a media buyer's head. When they leave, it goes with them.

So we made a simple change: before anyone launches a test, they have to log a hypothesis in Asana. What they're testing, what type of test it is (creative, offer, audience, structure), and one line on why they think it'll work. After the test, they come back and add the result and what they actually learned.

Each test gets its own task. Win, lose, or inconclusive, it all stays.

The goal is to build a database of our thinking, not just our results. Over time, AI can read every hypothesis we've ever logged and find patterns we'd never catch manually. Which bets tend to win. Which test types work on which platforms for which niches. The stuff that usually lives in one person's head and disappears when they leave.

We're a few weeks in. The habit is forming.

Talk soon,

Dmitry