The Chip Review Site I Waited 14 Years to Build

In the summer of 2011 I was driving from Missoula to Seattle, which is a lot of hours of I-90 and a lot of gas-station snacks. Somewhere along the way I went looking for an honest review of a bag of chips and came up empty. There were listicles, brand press releases, and the occasional forum thread. Nobody was taking chips seriously.

I filed the idea away. It stayed filed for fourteen years.

Then Christmas 2025 came around and I got bored. So I finally built it. That project is Chipter, and I've kept maintaining and growing it ever since.

A Silly Idea With a Real Content Problem

Chip reviews sound like a joke, and the writing is supposed to be fun. But once I sat down to build the thing, it turned into the same problem every marketing team I work with has.

A review isn't one piece of content. It's an article, a square card for the feed, a vertical video for Reels and Shorts, an animated preview for Bluesky, a newsletter, and a social post. Every one of them has to sound like the same opinionated narrator, whether it was written on a Tuesday in January or a Saturday in June. And Chipter doesn't have a content team.

So I didn't build a blog. I built a content platform: write the review once, and let the platform produce everything else. The case study has the full feature list and a technical tour. This post is about what I learned along the way.

The Voice Is the Product

The first thing I learned is that the fun part, the voice, is the hardest part to keep consistent.

The usual approach is to write a good prompt and paste it in wherever you need it. That works until you tweak it in one place and forget to tweak it somewhere else, and the newsletter starts sounding like a different person from the reviews.

What worked was treating the voice as content, not as a prompt. Chipter's voice lives in plain markdown files that get loaded into Sanity as reference documents the AI reads every time. There's a deep version for drafting whole reviews, a leaner version for quick edits inside the editor, and a laid-back version for when the house style is too much. Buttons right next to each field — Rewrite in Chipter voice, Generate verdict from review, Generate video copy — all pull from those same documents. Change the voice once, and everything follows.

I liked that pattern enough that I've since written about it for marketing teams: your brand guide should be content, not a prompt.

Automation Needs a Human in the Loop

The second lesson: automating everything is easy. Automating the right things is the actual work.

It would have been easy to let the platform publish whatever it produced. It's a chip site; what's the worst that could happen? But I wanted to see every video before readers did, and I didn't want an accidental double-post going out on Bluesky.

So nothing goes out on its own. When a video finishes rendering, it attaches to the review's draft, not the live page, so I see it first. The Bluesky post only fires when I hit publish. It's a small design decision, and it's the one I'd push hardest for on any client build. Marketing teams don't need fewer approvals. They need approvals that take seconds instead of a meeting.

Video Was the Hard Part

If the voice was the hardest part to get right, video was the hardest part to get working.

Rendering video in code is a wonderful idea right up until it has to run in the cloud, inside time limits, for formats every platform wants slightly differently. Chipter now has 15 video templates, from compact review cards to a papercut origin story of the potato. Getting there took real iteration.

What I learned there ended up being bigger than Chipter. It became the starting point for an open-source Sanity template I'll be writing about soon.

Side Projects Make Better Client Work

I didn't build Chipter to win clients. I built it because I wanted a chip review site and I was bored over the holidays. But it's turned into one of the most useful things I've built for my client work.

Client projects come with budgets, deadlines, and reasonable people who don't want to be the experiment. Chipter is where I get to be the experiment. When a marketing team asks whether their CMS can produce video, keep AI copy on-brand, or send its own newsletter, I'm not guessing. I've already built it, broken it, and fixed it on my own time.

It's also a reminder of something I tell clients a lot: a website is rarely the whole job. Chipter started as a review site and turned into a full-stack web app — accounts, comments, email, video, and AI — because that's what the content actually needed.

Still Growing

Fourteen years from idea to launch is a long time. The good news is the site doesn't have to wait that long for anything else. It's grown well past the Christmas version, and it keeps growing whenever I find another bag worth writing about.

Go read a review at chipter.io. And if your team has a content problem that's outgrown its website, let's talk.

I'm an independent full-stack web developer and Official Sanity Partner, working with marketing teams, design studios, and enterprise clients. Chipter is my own product: read the case study.