Reroute NJ: How a Pop-up Website Served 2,000 Commuters a Day

Hi, I’m media innovation journalist Ulrike Langer and you’re reading News Machines. This week’s post expands on a case study I featured in my keynote “Vibe Coding in the US Newsrooms: From Prototype to Strategy” at the AI for Media Network Meetup #8 in Munich on May 12. You can view and download my slides and the extended version of my speaker notes.

The case study is free as always. Paying subscribers get a Vibe Coding Field Kit as a bonus: a one-page Reroute NJ build sheet with Amditis's tech stack and costs, plus a decision checklist for editors weighing their own pop-up builds. Read on to the end of this post for the field kit.

Most conversations about vibe coding in the newsroom are about what gets built. The more interesting question, to my mind, is what happens after that.

Joe Amditis builds a lot of small interactive tools at the Center for Cooperative Media at Montclair State University in New Jersey. Reroute NJ is the one that has been getting the most attention. NJ Transit closed one track of the Portal Bridge between New Jersey and Manhattan earlier this year for four weeks of construction. Schedules changed. Commuters needed an alternative. Amditis noticed the information was scattered across TikTok, government websites, and local news posts. He built a station-by-station route planner in a day and a half, hosted as a public website with a defined endpoint.

Around 2,000 people used the route planner daily during the cutover. Three local newsrooms embedded it. The interface translated into ten languages. Then on March 16, when NJ Transit reopened the new track, Amditis added a banner saying Phase 1 was complete and stopped the daily scraper. The site is still up, but mostly dormant. He plans to reactivate it for Phase 2 this fall.

This is not a story about how fast vibe coding lets you ship. The build time is the headline; it is not the lesson.

Amditis has a term for the category. He calls it ephemeral software. He frames it as a new class that vibe coding makes possible: a website that exists for a defined window, then closes. "It can range from anything from what you would normally turn into a PowerPoint slide deck," he told me. "You might now turn that into an interactive website really quick, a quick GitHub page that goes up really quick for free that just has a little bit extra layer of interactivity and longevity”, Amditis told me.

The category is broader than what Reroute NJ does. A pop-up website can be a static information page about an election or a road closure. It can be an interactive tool like the one Amditis built. The defining property is the lifecycle, not the interactivity. The website is published, maintained while it matters, and shut down on purpose.

The pop-up restaurant comparison fits better than throwaway code. A pop-up has a known purpose and a known endpoint. It is temporary on purpose, not by failure. The quality of the food does not drop because the restaurant is closing in eight weeks.

This is different from a standard news article in a key way: the article is the default unit newsrooms organize around, even when it gets updated over time. A pop-up website is published, maintained while it matters, and shut down on purpose. The lifecycle is the product, not the byproduct.

Newsrooms actually have a precedent for this category: pop-up newsletters. The New York Times runs them around the Olympics. Bloomberg ran a Davos edition. The Information ran one around the Twitter acquisition. They have a defined audience, a defined endpoint, real editorial discipline, and then they wind down. Sometimes they come back. The audience accepts temporary as a feature, not a failure. Pop-up websites are the same logic, applied to a different medium.

A media tech vendor could have built Reroute NJ. None did. Amditis has a sharp answer for why.

The most useful feature the website offers is a PNG export. Commuters generate a card with their new route, save it to their camera roll, and never return. From a SaaS engagement perspective, this is a disaster. There is no return value once the user has what they need. No ads to serve, no subscription to upsell, no analytics to harvest.

"That return value is not very high," Amditis said. "There's once they have the value from the PNG or whatever, they don't need you anymore."

A grant-funded center of cooperative media can build that. A vendor whose model depends on recurring engagement cannot. The PNG-and-leave behavior is anti-vendor by design.

This matters beyond Reroute NJ. There is a whole class of public-service websites that fit the pop-up category: election-night turnout maps, flood-zone evacuation guides, road-closure trackers, school-closure dashboards. They serve a public, then go away. Vendors will not build them because the revenue model is wrong. Newsrooms can, if they have the staff capacity and the infrastructure to support a small builder.

The build was 1.5 days. The maintenance was several weeks, and that is the part the headlines leave out.

During the four-week cutover, Amditis was watching the site. When Governor of New Jersey Mikie Sherrill announced an early reopening, he had to update it. When Hoboken Terminal flooded mid-rollout and a second reroute kicked in, he had to push an alert banner: "There's some confusion. This may not be accurate until I can fix it." When NJ101.5 embedded the route planner and surfaced a variable that was not being sanitized properly, he had to fix it.

Amditis has no formula for how much maintenance to budget. "It's purely vibes based," he admitted. The closer he comes to a rule is a single question: is the personal time worth it, including maintenance? Not "can I build this," but "is this worth leaving in the world."

That question scales. A small newsroom asking whether to build their own pop-up site for an upcoming election or a planned road closure or a school district reorganization can ask themselves the same thing before they commit. The answer is rarely the number of hours to ship a prototype. The answer is the number of hours over the lifecycle, including the days when something breaks.

The category will only grow. Vibe coding makes the build cheaper every quarter. What does not get cheaper is the editorial judgment about what is worth building, the institutional support to pay for the tools and the time, and the discipline to retire something when its moment has passed.

Amditis has all three. He has a decade of experience with small publications. He has a $200-a-month Claude Max subscription that the Center pays for. He has a translation network of human reviewers, including the Center's Spanish-language coordinator Julie Delgado, who decide which words get translated and which stay in English because that is how they appear on the train station signs.

Most of this is invisible in the headline version of the story. The headline is "built in 1.5 days." The reality is one builder embedded in an institution that supports him.

Behind the paywall is a Vibe Coding Field Kit: a one-page Reroute NJ build sheet with Amditis's tech stack and costs, plus a decision checklist for editors weighing their own pop-up builds.