← Adam Garceau

Best Mileage Tracker App? I Asked AI and It Said Buy One

October 13, 2026 · The written version of my video. Reading time: about 7 minutes.

A while back I told you I was building a mileage tracker. I didn't build one. The tracker I ended up with is a free one somebody else already made, plus a small piece I built around it.

This post is the process, so you can run it on the next thing somebody tries to sell you. The short version is one habit: before you pay for software, or pay somebody to build it, open whatever AI chat you already have and tell it to argue against you. Not help you. Argue.

The problem

I drive for work: shoots, client meetings, and route days for a vending business I help run with my family. Every one of those miles is a deduction if you can prove it, and the IRS wants proof: the date, how far, where you went, why, and which business it was for. No log, no deduction.

My version is messier. I have two businesses, so every trip has to land in the right set of books. The vending business has a second driver who uses his own vehicle. And my QuickBooks subscription runs out in mid-October and I'm not renewing it. I already moved my books into a free program that lives on my own computer. So I needed miles to flow into those books without me remembering to open an app every time I got in the van.

Step 1: Write down what would make you regret it

I use a process I built with Claude called Forge. Before the AI writes anything, it has to research, write down what would make the project fail, and try to kill the idea. My first ask was huge: an app that replaces QuickBooks and tracks my mileage automatically. That is basically "build me a small software company."

It didn't start building. It asked: what would make you regret this in six months? That's the question to steal. I gave it three answers:

1. Missed miles. If it misses drives, I lose deductions.
2. Babysitting it. If I'm fixing an app every other week instead of making videos, that's a loss even if it works.
3. The books don't tie out. If my accountant can't use what it produces at tax time, none of it mattered.

I wrote those down before any research, so when I got excited later, the bar couldn't move.

Step 2: It told me to buy one

The AI looked at what already exists and said don't build this, buy one. It recommended TripLog, which has a free plan, with a paid plan at $4.99 a month. It also noted that tracking drives on a phone is genuinely hard, and the good apps have spent years tuning it.

For most of you, that's the right answer. One business, one car, you just need a log: get the free plan and be done in 10 minutes.

I didn't take it. It isn't the $5. It's that everything is $5. Five here, 14 there, QuickBooks on top, every month, forever, and you don't own any of it. Your data lives on their servers. What I'm really building toward is a small business that owns its own back office, on its own computer, with no monthly bill. The mileage app was just the first piece.

Step 3: Six skeptics came after me

A good process doesn't say "okay boss" when you override it. It went harder. It set up six separate AI reviews, each attacking a different part of the plan. Nineteen attacks, and six of them were kills, meaning the plan fails unless you fix it. In plain English:

• The phone shuts it down. The free tracking app I planned to use has a known problem: iPhone kills background apps, and users reported stopping after an hour to an hour and a half. A wedding day is longer than that. Tracking would have quietly stopped when I needed it most.
• Five things to maintain. The plan needed five separate pieces of software running on phones and a computer at home. One paid app is less work than five free things you keep alive.
• The paperwork. When you pay someone back for driving their own car, a clean log isn't enough. You need a written policy, which the IRS calls an accountable plan. (Spoiler: it turned out not to apply to us. My partner who drives the vending routes is the only owner of his business, so there's nobody to reimburse. His miles are a vehicle expense on his own return, and the trip log backs it up. If you pay an employee to drive their own car, you need it.)
• A person who has to use it. My second driver doesn't want to configure software, and nobody had put that on a calendar.
• My own promise. I'd said I'd share this publicly, but my real trip data would sit next to the public code. "I'll be careful" isn't a lock.

Step 4: Somebody already built it

It looked again and found Otograph, a free open source mileage tracker. Open source means the code is public and anyone can run it on their own computer for free. It works with a free phone app called OwnTracks (iPhone and Android) that sends your location to your own computer. Otograph turns that into trips, business or personal, with purpose, vehicle, and an exportable IRS deduction report. It uses the iPhone's own low-power location wake-ups, the ones Apple built for exactly this.

The person who made it is on GitHub, and I want to say thank you, because they did the hard part. Search "Otograph" on GitHub to find the project and give it a star. When I found it, it had two stars. If this post sends a few people their way, great.

It isn't perfect for me, and its own documentation says so: one person per install, no idea I have two businesses, tested on Linux. None of that was a reason to build my own. Here's how the process scored my options:

Option	Score
Build my own tracker	3 out of 10
Build only the small piece that's missing	7 out of 10
Build the small piece and tell the story, so other owners can do this	9 out of 10

Step 5: Build the bridge

Think of a bookkeeper between two desks: one holds the trips, the other holds my books, and the bridge carries the paper across. It sorts each trip into the right business, figures the deduction at the current IRS business mileage rate, and writes it into the right books. Anything it isn't sure about, a gap in tracking or an unlabeled trip, it sets aside for a person to look at. It doesn't guess.

I built it test first, on fake trips. The project has a wall for real data: a check that refuses to run if real data is in the wrong folder, and another that scans before anything gets saved. Careful is a habit. A lock is a lock.

Then I put Otograph on my home server with signups off and a nightly backup, and tested a restore, because a backup you've never restored is a guess.

What broke: getting my iPhone to talk to it took longer than building the bridge. OwnTracks can load settings from a QR code, a link, or a file, and all three failed on my iPhone. What worked was typing the settings by hand and pasting the long password from my Mac. My second driver's Android sent his location every two seconds and drained 58% of his battery in 8 hours. One setting fixed it. These were setup problems, not tracking problems.

The bridge isn't released yet. The code is private for now.

Run this on your own purchase

1. Write down what would make you regret it before you look at a single product. Three things max.
2. Ask the AI to argue against you. Say your office manager copies numbers into a spreadsheet every Friday and someone offers a custom dashboard. Ask: does the program you already pay for have a report that does this? Who fixes the dashboard when it changes? Sometimes the answer is a button you never clicked.
3. Ask who already built it. Most of your problem is already solved. Sometimes by a paid app, and $5 a month is fine. Sometimes by a free one.
4. Build only the leftover piece, the part specific to your business. That is where AI earns its keep, not in the flashy app.

If you want this set up in your business and don't want to fight the setup, comment with what kind of business you run.

Next, I let AI read a small business's spreadsheets and it found mistakes that had been sitting there for months: Claude Found the Excel Mistake That Overbilled a Customer.

I'm a marketer who builds his own tools. More at adamgarceau.com.