I Vibe Coded a Civic Tool

Everyone’s vibe coding the next todo app. The next AI wrapper. The next SaaS dashboard that solves a problem three other SaaS dashboards already solved.

I vibe coded a speed trap for my neighborhood.

It’s called Slow Them Down, and it’s a mobile app that lets you estimate vehicle speeds on your residential street using just your phone’s camera. Record a short video of passing traffic, mark two frames, get an instant speed estimate. No radar gun. No special equipment. No asking your city to do something they were never going to do without data.

Vibe Coding Has a Purpose Problem

I’ve been thinking a lot about the vibe coding discourse lately. I wrote about tokenmaxxing a few weeks ago — the idea that measuring AI value by volume of code produced is the same trap as measuring developer productivity by lines of code. The vibe coding conversation has a similar distortion.

The discourse is almost entirely about what you can build and almost never about what you should build.

A January 2026 paper titled “Vibe Coding Kills Open Source” argued that the practice weakens the open-source ecosystem by reducing meaningful engagement between users and maintainers. Maintainers of major projects like cURL and Ghostty have started closing external pull requests entirely because they can’t distinguish real contributions from AI-generated noise. The volume is up. The signal is down.

That’s a real problem. But it’s not a vibe coding problem. It’s a purpose problem.

When you lower the barrier to building software, you get more software. The question is whether any of it matters. The same AI tools that let someone spin up a throwaway SaaS in a weekend also let someone build a tool that helps their neighborhood collect traffic safety data. The technology is identical. The intent is not.

Privacy as a Design Decision, Not a Marketing Claim

I wrote recently about the security blind spots in AI meeting recorders which says that tools that capture every word in every meeting and ship it to someone else’s infrastructure. The core argument was that we’ve gotten so comfortable with the convenience of AI tools that we’ve stopped asking what data they’re collecting and where it goes.

Slow Them Down was built with the opposite instinct. All video processing happens on-device, there is no backend, there are no user accounts, there is no cloud upload, and there is no behavioral analytics. Faces and license plates captured in video frames are automatically detected and blurred on-device before you share anything.

When you’re building a tool that records video on residential streets, the privacy question isn’t theoretical. You’re capturing images of people’s cars, their homes, potentially their children. The only responsible way to build that is to make sure the data never leaves the device unless the user explicitly chooses to share it, and even then, it’s been scrubbed first.

The irony is that privacy-first architecture is actually easier to build in 2026 than it was five years ago. On-device ML models (Apple Vision on iOS, ML Kit on Android) handle face and plate detection without a network call. The processing power in a modern phone is more than sufficient. You don’t need a server. You just need to decide you don’t want one.

Open Source as Civic Infrastructure

The app includes a built-in directory of city, county, and state traffic departments so you can send your speed data directly to the people who can act on it. That directory is open source on GitHub. Don’t see your city? Open a PR and add it.

This is the part that connects to something bigger than one app.

The open-source model for civic tools is fundamentally different from the open-source model for developer tools. When someone contributes to a traffic agency directory, they’re not optimizing a build pipeline. They’re making it slightly easier for their neighbor to report a dangerous street. The contribution is inherently local, inherently specific, and inherently useful in a way that most open-source contributions aren’t.

I think about this in the context of the minutes added to workforce framing. AI tools are generating capacity — not just in engineering teams, but in individuals. A single person with an AI coding assistant and a weekend can build a tool that generates V85 speed calculations, produces PDF reports with distribution histograms, and emails them directly to a traffic department. That’s capacity that used to require a consulting firm and a city budget line item.

What do you do with the capacity AI gives you? That’s the question I keep coming back to. You can build another wrapper. Or you can point it at something that matters to your community.

The V85 and Why Data Changes Conversations

Here’s the thing that most people don’t know about traffic safety: cities don’t act on complaints. They act on data. Specifically, they act on V85 — the speed at or below which 85% of vehicles travel on a given street. It’s the standard metric traffic engineers use to evaluate whether a street needs intervention.

Before this app, getting V85 data required either hiring a traffic study firm or convincing your city to deploy speed monitoring equipment. Both take months. Both require someone in government to prioritize your street over every other street.

Slow Them Down lets a resident generate V85 data from their front porch. It calculates the V85, mean, median, and over-limit percentages. It produces exportable reports. It matches your location to the relevant traffic department and composes an email with the data attached.

The app doesn’t replace the city’s process. It gives residents the data they need to start the conversation.

That’s a fundamentally different model than “build it and they will come.” It’s “give people the tools to advocate for themselves and let the system work the way it was designed to work.” The system responds to data. Now residents can produce data.

What This Is Really About

I’ve spent 15+ years building engineering teams, scaling companies through IPOs and acquisitions, and implementing AI at the enterprise level. I’ve written about awe in an AI world — the feeling of watching technology expand what individuals and teams can accomplish.

Slow Them Down is the smallest thing I’ve ever shipped. It’s a mobile app with no backend, no revenue model, and a target audience of people who are frustrated about cars going too fast on their street.

It’s also the most personally meaningful thing I’ve shipped in a long time.

Because the question I keep asking myself isn’t “what can AI build?” We know the answer to that. The answer is: almost anything. The question is “what should the people who know how to wield these tools choose to build?”

I chose to build something for my neighbors. I chose to make it private by default, open source where it matters, and free. I chose to point the leverage that AI gives a single builder at a problem that affects real people on real streets.

That’s vibe coding for good. Not because the code was generated by AI — though much of it was — but because the intent behind it was civic, not commercial.

There’s more coming. I’m increasingly convinced that the most interesting applications of AI aren’t in the enterprise or in the next startup pitch deck. They’re in the spaces where individuals with new tools can solve problems that institutions have been too slow to address.

---

Slow Them Down is available on iOS and Android. The agency directory is open source and accepting contributions.

This post connects to ideas from Tokenmaxxing Is Lines-of-Code Thinking for the Agentic Era, Your AI Notetaker Is the Biggest Security Decision You’re Not Making, Minutes Added To Workforce, and Awe in an AI World.