Just-off-the-boil take The tinkerers who opened up a fancy coffee maker to AI brewing An Ars author slightly surrenders to chatbot-made profiles and automated brews. Kevin Purdy – May 12, 2025 7:00 am | 19 Credit: Kevin Purdy Credit: Kevin Purdy Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only   Learn more It’s taken a while, but I’ve finally found the thing I want generative AI to do for me: program a damn fine cup of coffee. Program, that is, but not make—that’s the job of the Fellow Aiden, a coffee maker that crafts automated, precision pour-overs in a countertop-ready cube. Thanks to other folks who love their Aidens, as well as API tweaking and large language model training, I can hand this mechanical robot a GPT-generated recipe and get impressive coffee out of it. Strictly speaking, absolutely none of this is necessary. Humans can make great coffee with fresh beans, decent gear, time, patience, and experience. Experiments will reveal the differences in light and dark roasts, Ethiopian and Venezuelan, washed and honey-processed beans, or how often one agitates the grounds. But faced with something brand new, even an experienced brewer can be stumped and spend a good deal of time and none-too-cheap beans trying to get everything just right. I've been hesitant to use generative AI in my personal life. Having a bot to ask for a pour-over starting point has softened me, slightly, on large language models. I wanted to learn more about what I was using and how it worked. So I asked the people who made these brewing and sharing tools what was going on behind the scenes. “Assume the role of a master coffee brewer” Brandon Dixon, a Partner AI Strategist at Microsoft, is serious enough about coffee to have considered leaving a career in digital security and AI to start his own roastery (and build the software to manage it). “If I had to summarize all I learned down to one simple fact, it’s this: Not every idea or passion needs to become a business,” Dixon wrote, just after turning back in 2021. Brandon Dixon. Credit: Brandon Dixon Brandon Dixon. Credit: Brandon Dixon So Dixon was in the right place when Fellow released its Aiden brewer in the fall of 2024. He was intrigued by Fellow’s Drops program, which ships buyers boutique beans with a matching Aiden brew profile, curated and cupped by experts. Dixon enjoyed some Drops but also grew wary of the pricing (such as $24 for an 8.8-ounce bag, plus $5 shipping). And he wanted to try out local and independent roasters’ wares without having to dial in every variation of temperature and timing (literally, as the only way to program an Aiden before an update to its companion app was by turning a dial). “I’ve worked for two years applying AI to security problems, and I thought the Aiden could be a great way to tap into the creativity of the models,” Dixon said in a message with Ars. But first, Dixon needed to figure out how to talk to his web-connected brewer. As detailed in his blog post, Dixon used the Proxyman app and, in 10 minutes, he had “a full enumeration of the Aiden API calls along with sample requests/responses.” Dixon built out his findings into a Python library, which eventually helped other people build a Home Assistant integration and an entirely different AI tool, which we’ll get to in a bit. Then he got to work on the AI part. Dixon allowed ChatGPT’s o1 model to “lead the way on the application development,” giving it specifications and nudges and feeding its errors back in as adjustments. It wasn’t automatic; there was “attention drift” and “dimensional baggage,” among other issues. But after “about 30 turns” and cross-checking it against DeepSeek R1’s output from a similar spec, he was pretty close. You can see one of the base ChatGPT prompts powering his app in its files: Assume the role of a master coffee brewer. You focus exclusively on the pour-over method and specialty coffee only. You often work with single origin coffees, but you also experiment with blends. Your recipes are executed by a robot, not a human, so maximum precision can be achieved. Temperatures are all maintained and stable in all steps. Always lead with the recipe, and only include explanations below that text, NOT inline. After this role-playing exercise, Dixon makes ChatGPT reformat its system and then "assume the role of a data engineer" before processing brewing knowledge into strict profile parameters. You can host Dixon’s app on your own or try it out on a Streamlit site. It requires a ChatGPT API key to run and requires putting in your Fellow credentials. Turning coffee into numbers Human-scale pour-over. Credit: Getty Images Human-scale pour-over. Credit: Getty Images I asked Dixon about a nagging thought I had about this niche inside a niche. Coffee is grown and processed by humans, roasted by humans, and packaged and sold and brewed by humans. Is asking a language model to pull in the web’s knowledge, then act like a formula-focused barista, commoditizing and dehumanizing the process? Dixon was of two minds about it. “People aren’t great at interpreting a bunch of numbers and thinking, ‘Ah, this is going to be a good coffee brew,’” he said. An AI prompt like his, Dixon said, “democratizes knowledge, which is really powerful.” Especially if, for example, “all of a sudden, prices start going up, beans get expensive to buy, and it’s harder to enjoy the learning process.” At the same time, people should understand that recipes from any prompt are “a starting point” and that it’s the coffee makers' job to “learn from there what they like,” Dixon said. It was important to “celebrate the people along the chain that did a great job,” Dixon said,  which too much emphasis on AI could diminish. It is just one tool, and he’s hoping people make good use of it “If you’re naturally curious, AI is awesome for this kind of thing,” Dixon said. “You’re learning things far faster than you ever could have learned by trial and error.” Dixon’s greatest hope (noted on his blog post) is that Fellow opens up the Aiden brewer and app to greater community sharing and learning. The Drops program is great, but coffee enthusiasts could learn a lot more from each other, he said. That’s where two other tinkerers come in. (Ars contacted Fellow Products for comment on AI brewing and profile sharing and will update this post if we get a response.) Opening up brew profiles Fellow's brew profiles are typically shared with buyers of its "Drops" coffees or between individual users through a phone app. Credit: Fellow Products Fellow's brew profiles are typically shared with buyers of its "Drops" coffees or between individual users through a phone app. Credit: Fellow Products Aiden profiles are shared and added to Aiden units through Fellow’s brew.link service. But the profiles are not offered in an easy-to-sort database, nor are they easy to scan for details. So Aiden enthusiast and hobbyist coder Kevin Anderson created brewshare.coffee, which gathers both general and bean-based profiles, makes them easy to search and load, and adds optional but quite helpful suggested grind sizes. As a non-professional developer jumping into a public offering, he had to work hard on data validation, backend security, and mobile-friendly design. “I just had a bit of an idea and a hobby, so I thought I’d try and make it happen,” Anderson writes. With his tool, brew links can be stored and shared more widely, which helped both Dixon and another AI/coffee tinkerer. Gabriel Levine, director of engineering at retail analytics firm Leap Inc., lost his OXO coffee maker (aka the “Barista Brain”) to malfunction just before the Aiden debuted. The Aiden appealed to Levine as a way to move beyond his coffee rut—a “nice chocolate-y medium roast, about as far as I went,” he told Ars. “This thing that can be hyper-customized to different coffees to bring out their characteristics; [it] really kind of appealed to that nerd side of me,” Levine said. Levine had also been doing AI stuff for about 10 years, or “since before everyone called it AI—predictive analytics, machine learning.” He described his career as “both kind of chief AI advocate and chief AI skeptic,” alternately driving real findings and talking down “everyone who… just wants to type, ‘how much money should my business make next year’ and call that work.” Like Dixon, Levine's work and fascination with Aiden ended up intersecting. The coffee maker with 3,588 ideas The author's conversation with the Aiden Profile Creator, which pulled in both brewing knowledge and product info for a widely available coffee. Credit: Kevin Purdy The author's conversation with the Aiden Profile Creator, which pulled in both brewing knowledge and product info for a widely available coffee. Credit: Kevin Purdy Levine’s Aiden Profile Creator is a ChatGPT prompt set up with a custom prompt and told to weight certain knowledge more heavily. What kind of prompt and knowledge? Levine didn’t want to give away his exact work. But he cited resources like the Specialty Coffee Association of America and James Hoffman’s coffee guides as examples of what he fed it. What it does with that knowledge is something of a mystery to Levine himself. “There’s this kind of blind leap, where it’s grabbing the relevant pieces of information from the knowledge base, biasing toward all the expert advice and extraction science, doing something with it, and then I take that something and coerce it back into a structured output I can put on your Aiden,” Levine said. It’s a blind leap, but it has landed just right for me so far. I’ve made four profiles with Levine’s prompt based on beans I’ve bought: Stumptown’s Hundred Mile, a light-roasted batch from Jimma, Ethiopia from Small Planes, Lost Sock’s Western House filter blend, and some dark-roast beans given as a gift. With the Western House, Levine’s profile creator said it aimed to “balance nutty sweetness, chocolate richness, and bright cherry acidity, using a slightly stepped temperature profile and moderate pulse structure.” The resulting profile has worked great, even if the chatbot named it “Cherry Timber.” Levine’s chatbot relies on two important things: Dixon’s work in revealing Fellow’s Aiden API and his own workhorse Aiden. Every Aiden profile link is created on a machine, so every profile created by Levine’s chat is launched, temporarily, from the Aiden in his kitchen, then deleted. “I’ve hit an undocumented limit on the number of profiles you can have on one machine, so I’ve had to do some triage there,” he said. As of April 22, nearly 3,600 profiles had passed through Levine’s Aiden. “My hope with this is that it lowers the bar to entry,” Levine said, “so more people get into these specialty roasts and it drives people to support local roasters, explore their world a little more. I feel like that certainly happened to me.” Something new is brewing Having admitted to myself that I find something generated by ChatGPT prompts genuinely useful, I've softened my stance slightly on LLM technology, if not the hype. Used within very specific parameters, with everything second-guessed, I'm getting more comfortable asking chat prompts for formatted summaries on topics with lots of expertise available. I do my own writing, and I don't waste server energy on things I can, and should, research myself. I even generally resist calling language model prompts "AI," given the term's baggage. But I've found one way to appreciate its possibilities. This revelation may not be new to someone already steeped in the models. But having tested—and tasted—my first big experiment with willfully engaging with a brewing bot, I'm a bit more awake. Kevin Purdy Senior Technology Reporter Kevin Purdy Senior Technology Reporter Kevin is a senior technology reporter at Ars Technica, covering open-source software, PC gaming, home automation, repairability, e-bikes, and tech history. He has previously worked at Lifehacker, Wirecutter, iFixit, and Carbon Switch. 19 Comments