I'm vibe coding beehiiv

Howdy wizards,

I’m doing it. I’m vibe coding my Email Service Provider, beehiiv (startup with ≈$30M ARR).

I’ve asked Claude to replace them for me to save costs.

Over 3 years, they’ve kept raising prices a lot.

My upcoming bill is $2,000/yr. At the rate my newsletter is growing, it’ll be $3,000 next year.

Thing is, they’re actually competitive on price.

That’s just the current market rate of sending a bunch of emails like I do.

I ran the math and over the next 3 years, I’m looking at:

This is for beehiiv’s cheapest paid plan, which still has lots of limitations.

Easily $10K in the next 3 years before any price increases. Over 5 years it’d likely be $20K+.

I’d been wanting to vibe code their platform for a while.

Earlier this month I had Claude rebuild my website in custom code, so I could ditch beehiiv’s website builder. It did it while I slept.

But my website is just a website. Replacing beehiiv’s email sending infrastructure, on the other hand, that’s a different beast.

I was concerned Claude would mess up. Or that this would turn into a huge time sink for me.

This was my brain before I realised it’s feasible:

Then last week I saw Marc Lou doing exactly it, vibe coding beehiiv, all by himself.

He’s one guy, not super technical as far as I know, vibe coding something that took a team of 100+ people to build.

I realised that, once again, I was in the familiar predicament of thinking too small when it comes to AI.

So I decided to ask Claude about this silly idea.

On a side note: do you deal with PDFs in your work sometimes? UPDF makes editing those unruly files quick and easy, and they have one hell of a deal for your Q4:

❦

I asked Claude to research replacing beehiiv so I could cut my operating costs.

It researched for an hour. That alone told me this wasn’t an easy task.

The response amazed me. These were the main points:

So, everything possible. At ≈95% discount. And it found a big thing to fix which beehiiv should have told me about.

I was sold.

Claude worked throughout the night. The next day I spent around 12 hours at my computer doing mostly other work. But also overseeing the building process a bit and making small decisions and authentications Claude needed, like registering to AWS.

Less than 24 hours after my first prompt I had a working email infrastructure and platform for my newsletter:

I estimate the build took 50% of my weekly token usage on my $200/month Claude plan. I used Claude Opus 5.5 on max reasoning throughout.

The platform has its own sending engine on Amazon’s servers, an editor modelled on beehiiv’s, open and click tracking on my own domain, A/B tests and much more. It also holds 3 years of readers’ history (over 1M records).

I now own my email infrastructure. Before I was renting the whole thing.

Now I’m renting just the top layer (the deliverability part). And I can swap that part out if they start hiking prices or go out of business.

Here’s what shifted in my brain once I started building this:

It feels like the software equivalent of people who move out to the woods and build a self-contained home. Moving off-grid, digitally.

A huge bonus is that this software will now be 100% personalised to me.

I already have an app for designing cover images for this newsletter. And I’ve been keeping a separate spreadsheet with a content calendar, sponsor tracking, subscriber acquisition statistics, and more.

That all went straight into my new newsletter dashboard. One home for everything related to my newsletter, and nothing I don’t need.

It’s also agent-first. beehiiv’s platform doesn’t enable every action through their API. Which used to mean me having to click into their web interface myself.

With my own custom-built platform, Claude can run every action on my behalf. Drafting, sending, automating, analysing. You name it.

Everything is clay and can be moulded to my preferences.

And my preference is giving Claude access to absolutely everything.

I can finally be the director and tastemaker. Instead of keyboard typer and mouse clicker.

I’m still going to be on beehiiv for a little while. Because my new setup is a new delivery route, it means I need to warm my sender reputation gradually to avoid getting flagged as spam. Technical stuff I didn’t know about before, but Claude explained it all to me.

So I’ll do a gradual migration over 1-2 months. I’ll send a fraction of these emails through my own sending engine at first, then a little more each week, until I reach the full size of my subscriber list.

That means I have not tested the most tricky parts of this at scale yet.

I’m quite confident it’ll work, but I’ll report back whatever the outcome.

The usual pushback to projects like this: building is the easy part, and maintaining it becomes your new full-time unpaid job. I agree the work doesn’t stop at building, but it won’t be my job: AI already maintains a lot of my automations on a schedule from my own VPS (shout-out to Hetzner), and alerts me when something breaks.

My thesis is that maintaining this will be like another build, just run on a schedule.

Also — my situation is probably different from yours. I run a one-person business, so if this breaks, I’m the only one paying for it.

If you’re in a bigger company, where you can’t just vibe code mission-critical software: don’t let posts like this give you FOMO.

Being the crash test dummy and doing this kind of stunt is kind of my job running a newsletter about AI.

❦

You are a delight.

Dario

Disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes work with sponsors and may earn a commission if you buy something through a link in here. If you choose to click, subscribe, or buy through any of them, THANK YOU – it will make it possible for me to continue to do this.