Eat the Damn English Muffin

“Eat the damn English muffin.”

That was, more or less, the advice ChatGPT gave me that morning.

This was not where I expected the conversation to end. I had asked it to find me a keto-friendly breakfast in downtown Seattle.

About twenty minutes earlier, I had made a slightly provocative claim:

“I think you can replace Yelp. Find me a killer breakfast spot.”

Its first answer was Ludi’s.

That was a good answer.

It was also a place I already knew and loved.

So I changed the problem.

I wanted something quicker.

ChatGPT suggested a few alternatives. Then I added another constraint: make it keto-friendly, make it close, and make it somewhere I could walk to without turning breakfast into an expedition.

The recommendation changed again.

This time it sent me to Mainstay Provisions in Denny Triangle.

A recommendation you can argue with

Mainstay has a breakfast sandwich called the McMainstay: fried egg, house-made sausage, American cheese, spicy aioli, and a house-made English muffin.

There was an obvious solution to the carb problem.

Lose the muffin.

ChatGPT even had an order ready for me: ask for the egg, sausage, cheese, and aioli in a container.

So I started walking.

This was already different from the usual restaurant-search ritual.

Normally I would open a map or Yelp, search for breakfast, look at stars, scroll through photographs, open a few menus, read some reviews, and eventually choose something.

Instead, I kept changing the question.

Good breakfast became something quicker.

Quicker became something close.

Close became something keto-friendly.

And every time I changed the problem, the recommendation changed with it.

Then I got to Mainstay and looked at the menu.

The English muffin was house-made.

That complicated things.

I sent ChatGPT a photo of the menu.

Then I asked the question that destroyed the carefully constructed dietary requirements of the previous several minutes:

“Or I could just have the carbs. What do you think?”

The answer, essentially, was:

Eat the damn English muffin.

This seemed reasonable.

The sandwich

Only a few people were seated. A couple stood at the counter, studying the menu, apparently trying to decide what to order.

Clearly, they hadn’t asked ChatGPT.

They stepped aside, still contemplating their options. I walked up to the counter knowing exactly what I wanted.

The person taking my order was cheerful. I ordered the McMainstay and added bacon.

So much for keto.

Then came the question:

“For here?”

I thought about it for a second.

This sandwich deserved a plate.

So I stayed.

Mainstay was crisp and clean, with a handful of people scattered around the café. I found a seat, settled in with the coffee I’d brought from earlier that morning, and got absorbed in email on my phone.

Then I heard my name.

Someone was carrying out my McMainstay on a plate.

“Added bacon—nice!”

I looked down at the sandwich.

The sausage patty was enormous. Not your typical fast-food breakfast sausage. This was nearly a burger patty.

And the bacon was legit.

Thick. Crisp. Exactly what I’d hoped for.

I knew better than to attack it immediately. A couple of minutes to cool down would pay off.

And it did.

The sausage was delicious. Somewhere between it and the bacon was a subtle sweetness. The egg was cooked perfectly, the American cheese had melted into everything, and the house-made English muffin—the ingredient I’d originally planned to eliminate—absolutely belonged there.

I quickly discovered that this wasn’t a sandwich I could casually eat with one hand.

It required two.

It was messy.

It was delicious.

By the time I’d finished, something had changed.

The quiet little café I’d walked into was suddenly packed. The number of people seemed to have quadrupled while I was occupied with breakfast.

Maybe I’d timed it perfectly.

Maybe I’d just beaten the rush.

Either way, I was glad I’d stayed.

Maybe the list was the problem

I don’t actually think Yelp became obsolete because I ate a good breakfast sandwich.

Restaurant reviews solve a useful problem. So do maps, star ratings, photographs, menus, and lists of the best breakfast sandwiches in Seattle.

But sitting at Mainstay, I realized I hadn’t really used AI as a better restaurant database.

I had used it differently.

I didn’t know exactly what I wanted when I started.

I knew I wanted breakfast.

Then I remembered I wanted something quick.

Then I decided I should eat keto.

Then I decided it needed to be close.

Then I arrived and discovered that the bread I was planning to remove was made in-house.

Then I changed my mind.

Again.

That’s much closer to how people actually make decisions than the filters on a restaurant search page suggest.

We don’t always begin with a cuisine, price range, neighborhood, and desired star rating.

Sometimes we begin with:

I’m hungry.

And figure out the rest along the way.

The interesting thing about conversational recommendations isn’t that an AI can tell me where restaurants are. We’ve had databases capable of doing that for decades.

It’s that I can argue with the recommendation.

I can add a constraint. Remove one. Tell it I’ve already been somewhere. Show it the menu. Ask what looks good. Change my mind. Ask whether the thing I’m about to order defeats the entire reason I chose the restaurant in the first place.

I don’t know whether ChatGPT can replace Yelp.

But I know something more useful than I did that morning.

Sometimes the recommendation isn’t the restaurant.

Sometimes it’s knowing when to change your mind.

Eat the damn English muffin.