The Little Blue Menu and What It Already Knows About AI
There’s a Chick-fil-A in College Park, Maryland, that doesn’t look like a Chick-fil-A. They painted it blue instead of red. The menu…
The Little Blue Menu and What It Already Knows About AI
Crossing the Potomac
There’s a Chick-fil-A in College Park, Maryland, that doesn’t look like a Chick-fil-A. They painted it blue instead of red. The menu carries burgers, bone-in wings, sweet potato tots dusted with Old Bay, and for a while, it ran five different pizzas, including one topped with chopped chicken nuggets and pickles. It opened in September 2023 at 7242 Baltimore Avenue, in a building that used to be an Applebee’s. They closed it for a remodel in May 2025, reopened in August with dine-in seating for a hundred, and it’s still the only location in the country serving any of it.
The place is a test market. Chick-fil-A has been running test markets since Truett Cathy was changing items on a blue paper menu at his original lunch counter in Hapeville, Georgia in the 1940s, working out which sandwich people kept coming back for. The College Park location is named after that menu. The pattern itself is older than any of us, and when a quick-service chain runs one nobody calls it innovative, because everyone in retail already understands why you’d prove a concept in a single ZIP code before committing to fifty thousand of them.
The pattern stuck around because building a real-world prototype was expensive. A lease, a kitchen, a staff, a supply chain, a permit. Someone had to figure out whether a pizza would survive a 10-minute delivery in a hybrid sedan without going soggy. Failure had to be quarantined so it didn’t take the brand down with it. So you picked one location, near a university where the demographic skewed young and curious and willing to drive an extra fifteen minutes for a chorizo pie, and you let people taste the chicken-nugget pizza and tell you whether the proposition was inspired or insane.
In software, the equivalent process used to take months. The cost of prototyping has dropped to almost nothing. A working build that needed a small team and a sprint cycle two years ago now comes together over a weekend on a laptop, with the state management, the accessibility metadata, and most of the boring edge cases roughly handled. The kind of friction that used to keep half-formed ideas trapped at the whiteboard has mostly gone away.
What this means in practice is that the failure rate for ideas is no longer the bottleneck. The bottleneck has moved. It used to live in capital and engineering hours, both of which were scarce enough to support an entire venture industry built around rationing them. Both are abundant now, or close enough. What’s scarce is the ability to look at ten shipped prototypes and tell which one of them is actually worth pursuing.
For the builder, this is mostly an unambiguous improvement. The math gets better. Run twenty experiments instead of two and your odds of stumbling into something people actually want go up. This is the same bet Chick-fil-A is making in College Park, which is to say it has been the right bet for retail since the 1960s.
When the cost of execution drops to near zero, the cost of attention goes the other direction. Every prototype that ships is one more thing competing for the same finite supply of eyeballs and the same fixed number of hours in a day. The signal-to-noise ratio collapses from the noise side. The market wasn’t quiet to begin with, and now everyone with a laptop and an API key can add their own pilot to the pile, so the pile gets taller every week.
The old reason marketing mattered was leverage on sunk cost. Two years and three million dollars went into building a thing, so another half a million had to go into telling people about it, because the alternative was eating the whole loss. Most of that marketing existed to recoup the bet already placed, not really to sell the product itself. The marketing function inside most companies was a hedge on the engineering decision, dressed up in different language for the org chart.
That equation has flipped. Execution is cheap, the bets are small, and the channel through which you reach a customer is more crowded than it has ever been in the history of channels. There is no AI tool that grants your specific product the right to be heard over the next person’s. The harder problem now sits downstream of the build.
I think this is what most people miss when they talk about AI eating software. They model the supply side correctly, that anyone can build anything, and then they assume the demand side scales with it. It doesn’t. Human attention is fixed by the same number of waking hours it had in 1960, and the channels that mediate it (the App Store, the algorithmic feed, the search result, the inbox) are gatekept in ways that no faster prototyping cycle can route around. Shipping faster doesn’t get you any closer to being heard.
The College Park location is instructive on this front, too. Chick-fil-A didn’t just build the kitchen. They picked College Park specifically. The University of Maryland sits a mile down Route 1 from the front door. The student body provides a built-in flow of curious diners with high turnover, low brand loyalty, and a media presence on TikTok and Instagram that does the marketing for free. There’s a reviewer working under the name Mr. 8 That, a Montgomery County math teacher who rates restaurants on a 1-to-10 scale, and his Little Blue Menu video has thousands of views. Chick-fil-A didn’t pay for that. Their food testing and their social listening were happening in the same room, on the same nights, with the same people.
If I were starting something now, I would spend a lot less time worrying about whether I could build it. Of course, I can build it. I would spend much more time worrying about where the College Park is. Where is the geography, physical or digital, that gives a small bet enough surface area to attract attention without burning every available dollar on cold acquisition. Most of the AI-era founder discourse I read seems to assume the answer to that question takes care of itself, which I don’t think it does.
Test markets were never really about testing the food. The food was easy enough to fix. The point was finding a place where a small number of people would tell you what was wrong before you had to find out from a much larger number of people in a much more expensive way.
Anybody can stand up the kitchen now. Fewer people are thinking carefully about who they’re cooking for, where those people already gather, and how feedback travels back from the diners to whoever’s holding the spatula. Chick-fil-A has been thinking about that since before I was born. They reopened the College Park location in August 2025, with more seating and a refreshed menu, which tells you something about how durable the underlying instinct is. There are ten thousand AI-built prototypes shipping this week, and most of them will fail for reasons that have nothing to do with the code.
By Joshua McDonald on May 5, 2026.
Exported from Medium on August 26, 2026.
Reader discussion