About Squeezed

I'm the guy my friends call before they buy a used car.

No favor owed, no waiting for me to text back. Type the car into Squeezed, or paste the VIN, and it tells you what I would over the phone: what breaks on that exact car, what the fixes cost, and what to ask before you put money down. In 45 seconds.

Keep reading Try Squeezed

15+ yrsProduct & UX Design

PrincipalProduct Designer

LifelongBMW Enthusiast

I've spent over 15 years as a Principal Product Designer, leading UX at startups, AI companies, and e-commerce brands — building products that handle massive complexity without making people think about it. I've spent just as long buying and selling cars on the side, for the same reason: I like taking something complicated and making it obvious.

Current project

2000 BMW M Coupe — 185k miles

Old BMWs are my thing. Finding the good ones, keeping them on the road, knowing exactly which ones to walk away from. I've bought and sold enough to know the difference between a gem and a money pit — which is exactly what Squeezed helps you do.

Follow the build on Instagram

The Origin

It started with texts from friends.

A text from a friend

A friend texting a used-car listing link

It always started the same way: a link to a listing and a 'what do you think?' That one question meant pulling the exact year, trim, and engine, then figuring out what actually goes wrong on that specific car: the known failure points, the recalls, the repair bills that turn a good deal into a bad one. The kind of thing you only know if you've looked at hundreds of these.

Overnight research guide

Instead of reading ten thousand forum posts by hand, I put AI research models to work on the noise: forums, NHTSA recalls, service bulletins, owner complaints. I checked the findings, cut what didn't hold up, and built each one into a single guide a friend could open on the test drive. The research was the AI's job. Knowing what mattered and what to trust was mine.

Squeezed, at scale

A Squeezed report open on a phone at a car lot

Squeezed is that exact process, productized. The same research models, the same eye for what's actually worth checking, the same one-question prompt: year, make, model, mileage. What used to cost me a free Saturday now takes 45 seconds, and it works for any car, for anyone, without waiting on me to text back.

I've always been the guy my friends call before they buy a used car. Not because I catch problems — because I help them find the gems.

A friend would text me a link. I'd put AI research models to work on it — forums, NHTSA recalls, service bulletins, Facebook groups — then go through what came back, throw out the noise, and build the keepers into a custom buyer's guide for that specific vehicle. They could open it on their phone, take it to the test drive, and check everything I'd flagged.

I kept refining those guides. More data points, better sources, sharper questions. Every year they got better because every car taught me something new.

Detail-oriented, research-driven. That's how I design products for a living, and it turns out it's exactly how you should buy a car.

At some point I realized these two parts of my life were solving the same problem. The guides I built for friends — AI doing the research, me deciding what mattered — that was the product. It just needed to exist at scale, for anyone, for any car, without waiting on me to have a free Saturday.

The Garage

I have a problem. I buy old cars.

Thirty-five of them so far, and not one bought new. I refuse to. Anything under four or five years old is somebody else's depreciation, and I would rather have theirs than mine. So I hunt the bottom of that curve: the point where a car has already lost most of what it is going to lose and is finally worth what it costs.

Which means thirty-five sets of high miles, patchy service history, and sellers who were not always telling me everything. Thirty-five rounds of working out what goes wrong at what mileage, what a fix actually costs, and which problems are deal-breakers versus which are just a Tuesday. You learn fast when it is your own money.

Squeezed is that, pointed at whatever car you are looking at. Every hard lesson I paid for, before you pay for yours.

A slice of the collection. Mostly BMWs, because that is where the problem started.

Right Tool, Right Job

LLMs are good at one thing: finding signal in noise.

Most products stuff “AI” in because they’re expected to. This is the opposite. We use LLMs for exactly what they are: pattern-matching engines that can ingest mountains of messy, noisy data and pull out the signal.

Forum posts, service bulletins, recall databases, owner complaints across three generations and six trim levels — that's what LLMs are good at. Making things up, being creative, or replacing human judgment? That's what they're bad at. So we only ask them to do the first thing.

Squeezed doesn't predict whether a specific car will break. It tells you what's known to break on that year, make, model, and mileage — backed by documented data with confidence levels on every finding.

That's a research assistant that can read 10,000 forum posts in 45 seconds. Not magic — just good engineering.

Documented, not guessed

Every finding is sourced and confidence-rated. No black-box scores.

No prediction games

We tell you what’s known to fail on that car. Not what might.

You make the call

Squeezed gives you ammunition for the test drive, not a verdict.

Why No Chatbot

One search box. Nothing to prompt.

Everyone wants to bolt a chatbot onto everything right now. We use research models for the part they're actually good at, reading mountains of forum posts and recall data, and skip them everywhere else. Getting your car into the system is a search box: type “2014 Ford Mustang”, or paste the VIN and we decode the rest.

It looks like a chat box and behaves like nothing of the sort. There is no prompt to phrase and no model guessing what you meant: the box matches what you type against 70,000 real vehicles and gives you the exact one to pick. From there it is a handful of labeled questions, trim, mileage, what you care about, each one visible and exactly where you expect it.

The research models do the hard part: reading thousands of forum posts so you don't have to. The interface does the easy part: getting your car pinned down exactly, in seconds, without making you think about it.

Type it the way you say it

“2014 Ford Mustang” is enough to start. The box matches as you type and hands you the real vehicle to confirm. Already have the VIN? Paste it and we decode year, make, model, trim, and engine for you.

Bad input, bad report

Your report is only as accurate as what you enter. A mileage field rejects “15ish thousand”. A model dropdown can't be misspelled. Garbage in, garbage out — so we don't let garbage in.

The same flow, every time

No prompt to write, no wondering what to say. The questions are the same ones, in the same order, every time you run a guide.

Do The Math

$0 vs. $2,000+

$0Your First Guide

$3.99Every Guide After That

$150–300A Pre-Purchase Inspection

$2,000+A Bad Buy’s First Year

The first five are on us, and after that it is pocket change before you spend thousands. Squeezed doesn't replace a mechanic — it replaces the 10 hours of panic-research you'd do the night before you go look at a car.

Not because you don't know what you're looking for — because there's too much information to sort through alone.

Stop gambling on used cars

Get smart before you buy.

One guide. 45 seconds. Every known failure pattern, repair cost, and question to ask the seller.

Guides squeezed so far

Try Squeezed Browse Guides