How This Got Out of Hand
Ian Lang ·
I wasn’t always into horror. In fact, it used to not make sense to me—why would someone want to intentionally feel scared, or uncomfortable? Why would you want to see grotesque imagery? I have (obviously) since come around, but until a couple of years ago, my wife was the sole household horror fan, watching stuff on her own or when I was out of town. And that was fine; she could have her thing, and I could have mine.
But then a couple of Octobers ago, something changed: I sat down and watched some, as a spooky season concession to her. Friday the 13th, to be exact. Nothing flashy or groundbreaking by today’s standards. Not even an especially good horror film, if we’re being honest. But it worked. It started to make sense. This was fun. This was an unusually deep, diverse genre. You can say a lot of things via horror. You can also say nothing at all and simply use it as a vehicle for gratuitous nudity and absurd deaths. It contains multitudes.
Of course I wanted more. A lot more. Like, every night and sometimes twice more. I found titles the way I imagine most people do: IMDb/RottenTomatoes ratings and lists. Online “best horror” guides. Endlessly scrolling streaming platforms and hemming and hawing over titles for longer than the eventual movie’s runtime. Saying “eh, this poster looks pretty cool, right?” and then watching something that turns out to be mega, turbo ASS. And all that worked—for a while.
I wasn’t always an AI user, either. Like many, I derided early LLMs as little more than resource-intensive auto-complete software—because early on, that’s functionally all they were. But about a year into it, ChatGPT had become useful enough personally and professionally to earn a spot on my homescreen. Naturally, I thought excellent, I can use the expensive clanker on my phone to find more horror movies!
And it worked! Save for two issues: The Autopsy of Jane Doe and The Taking of Deborah Logan. Both fine movies. Superlative, even. But as humans, we instinctively understand that recommending someone the same two movies—especially after they’ve told us they’ve seen them—would come off as insane. AI is very good at taking natural language and translating it into action. It is decidedly less good at actually knowing or remembering anything the way a person would. So here I was, night after night, giving it fairly specific requests, yet being recommended those same two goddamn movies. Artificial “intelligence,” my ass.
So I said: Fine. I will keep track of what I watch, and I’ll do it in the cloud so the machine has a live record. That, however, didn’t stop it from recommending:
- Stuff I’d seen and hadn’t remembered to tell it
- Stuff that lived on platforms I could not access, or no platform at all.
So I said: That’s fine too. I’ll just tell it what platforms I do have. Which, likely because something is very much wrong with me, turned into hey, why don’t I just start cataloging every horror title on these platforms, adding meta data, and keeping them all in a live spreadsheet so ChatGPT never recommends me something I can’t access or have already seen?
Thus began what would become this site. At first I was using ScreamingFrog to scrape sitemaps, manually updating and categorizing titles by the hundreds. Then one morning OpenAI released ChatGPT Work, and it was like a lightbulb went off: I could suddenly, systematically and at scale, scrape thousands of titles for the narrative data that matters to me. Alongside it I built a taste model that logs what we watch and what my wife and I like and dislike about each title. ChatGPT would be the gateway, but my ludicrous spreadsheets would be the lone source of truth.
It worked wonderfully, and it’s how I personally still find new movies to this day. But one recent Monday, I was smoking a cigarette on my back porch when I thought why doesn’t everyone have something like this? A week later, I had v1.0 of the website you’re currently on. I learned about things called “Cloudflare,” and “query contracts,” and someone named “JSON.” I tried, mightily, to build it so that no AI involvement was needed, only relenting when I realized the tool would never work very well without using AI to translate normal language into something the reviewer could use to query my database.
The strangest part has been watching other people use it. I assumed a few fellow horror weirdos might find it interesting, but instead I've had strangers telling me about the movies they've discovered, the ones they've watched, and the ways I've managed to break my own search engine. It's been equal parts gratifying and terrifying.
It’s far from perfect. And I have no idea what, if anything, this will ultimately become. But it’s here, it mostly works, and you have 7300+ obsessively categorized horror titles to find. Maybe eventually I’ll expand the corpus. Maybe I’ll eventually sync live streaming info. Maybe I’ll one day sell it to some big streaming platform who will make it shitty and unusable because they only want the data. I genuinely have no idea. But I know that it’s a labor of love, and I know it can help you find the right horror movie for tonight.