Solve It Once and Never Again with Damien Retzinger of Graycore

I have known about Daffodil for a while and I did not understand what it was for until I built a Magento store with it in about thirty minutes and had it live inside an hour. Damien Retzinger is the founder and software engineer at Graycore, and he describes himself as the BDFL of Daffodil, an open source ecommerce front end framework. He came on to talk about where it came from, how he writes code now, and what he thinks happens to Magento in the US.

Eighty Hour Weeks in an Attic

Damien started on Magento at twenty two, right out of university, at a two-person company that put him in their attic to build a store for a customer.

“I didn’t know Magento at all,” he says. “There were very little tutorials on the internet at the time. So I spent a silly amount of time reading code and failing that first summer into fall. I think I was pulling like 80-hour work weeks for a solid six months. But by the end of it, I had a working store.”

My own start looks the same from a distance and worse up close. I built a thing called Cardex on Magento 1 that added a product to your cart when you bought something else, and I made a lot of money modifying it for people who each wanted it a little different, and I never captured the reusable part. Damien did.

The Migration That Produced Daffodil

Daffodil came out of the Magento 1 to Magento 2 transition around 2016 and 2017. Damien had merchants who wanted to upgrade and who did not want their customers to learn a new website because the system underneath had changed.

“So Daffodil came out of what if I could rip the front end off, make it look the same in Magento 1, and then keep, write an API and then just swap the API up for Magento 2’s API,” he says. “I built a driver layer between the front end and the back end. I made an API for Magento 1, and then I made an API for Magento 2. We launched it on Magento 1, switched it over to Magento 2, and now the user interface was the same.”

He still rebuilt the whole thing from scratch that first time. The difference is what he told the merchant.

“I’m willing to take on that cost because for me, I want to solve this problem once and never again.”

His analogy is a printer. You buy one, plug it in, and it works, and ecommerce has not worked that way for a developer. “Even if you know React or Angular or Vue or HTML and CSS, you don’t necessarily know how to build an e-commerce store,” he says. A front end developer who moves platforms loses the whole skill set, and he wanted that problem gone.

Daffodil has about 60 libraries, one per unit of ecommerce, products, search, categories, cart, checkout, orders, customers, each with an interface that describes how the thing has to behave and a standard driver against it. It connects to Magento and Shopify, a Medusa driver is in progress, and in his words, as long as there is an API, GraphQL or REST, they can wire it together. It is open source, and he reviews contributed drivers himself.

Code Review Is the Job Now

Damien started coding with AI about a year and a half ago, asking ChatGPT whether code he had written had problems. That is not the workflow anymore.

“These days actually I find that I am primarily doing code review,” he says. “I would say probably 75, 80 percent, maybe even 90 percent of what I’m writing isn’t handwritten by me. The ideas are mine, but I’m letting the robot write the code and then I’m reviewing and tweaking. My acceptance rate isn’t particularly high. Most of the time I hate what it writes.”

He plans before he lets anything rip. He recommends a Claude skill from Matt Pocock called Grill Me, which interrogates you about the problem before a line gets written, rubber duck debugging without the duck.

I pushed on the piece of this that worries me, which is people hitting enter because they want it to go. I do it myself, and in marketing I think that race to mediocrity is going to hurt. Damien widened it past marketers.

“The people who are gonna stand out are people who are putting their ideas into the system and not just hitting yes, as if it’s a slot machine,” he says. “Really these things are amplifiers and they’ll amplify the bell curve. If there’s a lot of mediocrity in people, it’s gonna amplify a lot of mediocrity. However, for those individuals who are at the one end of the bell curve, I would expect it to be a really massive amplifier for their opportunities.”

On what that does to a developer’s pay, he was blunt. Lines of code was always a terrible metric, and features per hour is a different one, and if you close more tickets you get paid more. My son is a front end developer who has embraced these tools completely, and I think the developers insisting on writing every line by hand are going to get paid less per hour. Damien agreed.

His two-year read is that the individual contributor becomes a reviewer. “Most of what your day is going to be is reading code and not necessarily writing it anymore,” he says. The other half is business users vibe coding, and somebody cleaning up after them. “There are definitely going to be companies that solely offer code review for vibe coded applications.”

Headful and Headless

Damien coined a word during the conversation, headful, for a Shopify store running Shopify’s own front end, and I am keeping it. His view on the monolith versus headless argument is that it ebbs and flows, and that specialization wins in the end because people lean front end or back end and a headless theme lets them stay on their side of the network barrier.

He also does not think Daffodil is for everyone. “If you are looking for something fast, Shopify, pick Shopify,” he says. “If you want to do something that’s more longer lasting that you have control over, this is where Daffodil shines. It’s where Magento shines, it’s where open source generally shines. You can express yourself creatively through open source.”

Asked whether Daffodil connects to Adobe Commerce Cloud as a service yet, he said he has asked for access numerous times and has yet to hear back. I have been hard on Adobe. That one might be the project that brings me back.

The Magento Lull

Both of us are in the US, and I have watched the energy around Magento drain out here while it holds in Europe. Why can nobody here get behind it anymore? Damien’s read is that Magento has always been difficult, that Varien, eBay and Adobe never solved that, and that the ecosystem is heading into a quick lull. He thinks AI ends it.

“A junior developer can jump into a code base today that they have no idea what’s going on, but they know how to code and they know some of the principles of software engineering,” he says. “They can jump in and play around and break stuff and try things. They can submit PRs a lot faster than they used to be able to before. I think that alone is a huge positive for anything like Magento, anything like these large complex code bases with high barriers to entry.”

One Click

The plug at the end is the thing I had already used. Damien spent a year trying to make the barrier to Daffodil zero, then realized a front end with no back end is not much use, and the developers willing to write code are Magento developers. So he turned his own ten-year-old Docker Compose setup for Magento into a GitHub Codespace, added the choices people care about, Hyva, MageOS, MageOS minimal, Daffodil, added his CI pipelines, and added Claude.

“I think single one button click, make it possible, make it entirely open source, make it entirely free, and get people started,” he says. “I think that’s what I try to solve. And that’s what I think I did solve.”

It is free for thirty hours a month of Codespace time. Someone has already sent him a B2B starter kit to add as an option.

“I think the best thing any e-commerce company can do is figure out how to make their offering one click,” he says. “I’ve been building Daffodil for ten years and it took me ten years to figure out that answer.”

The starter is at ecommerce-ai-starter.graycore.io. Damien Retzinger is on LinkedIn, Graycore is at graycore.io, and Daffodil is on GitHub.