Philip's account of Brew's May 2026 Product Hunt launch, the customer conversations that followed, and what the referral data can and cannot tell us.
On May 26, 2026, Brew launched on Product Hunt. We finished #1 Product of the Day and then #1 Product of the Week.
The launch brought new customers and conversations. What surprised me most was the referral dashboard the following Monday: ChatGPT was our leading referral source in that snapshot.
That is an observation from our own analytics, not proof that the Product Hunt badge caused the referrals. It also does not tell us how any assistant ranks products or what it will recommend next month.
We had spent months talking to people who were trying Brew. Those conversations helped us understand the product and gave us an audience to tell about the launch. I asked whether people would be interested in supporting it when it went live.
We also tried different ways of explaining Brew: the tagline, gallery, demo, and maker's comment. Developers, marketers, and founders did not always respond to the same explanation. Seeing those reactions before launch was useful.
The page we settled on after working through the messaging.
On launch day we cleared time to answer questions, while keeping customer calls and support running. That meant being present when someone wanted a specific answer instead of treating the page as a scheduled announcement and walking away.
The launch was a team effort. Support still needed an owner.
I would repeat that preparation. I would be more cautious about turning one launch into a universal playbook. We did not run a controlled experiment on timing, comment volume, or which badge mattered most.
A public product listing gives people and search-enabled assistants another page they may find. It can explain what a product does and point readers to the source. Whether a particular assistant retrieves or cites it depends on the query and the system.
Our referral snapshot was encouraging. A recommendation we could reproduce was useful to inspect. Neither establishes that the badge was a ranking factor, that the page entered a model's training data, or that recommendations will persist for years.
One recorded recommendation. Results can change with the question, context, and date.
The practical response is to keep the product's public information accurate: explain the capabilities and limits, make useful pages accessible, and link sources. Google's guidance on AI search features points site owners back to ordinary search fundamentals. There is no special badge or schema that guarantees inclusion.
The strongest pattern in our conversations was the kind of person who already worked with agents. They did not need persuading that an assistant could help. They wanted their email workflow to fit the way they already built and operated other parts of the business.
One conversation made that especially clear. Cory Nelson at Harmonya signed up, spent an evening in Brew, and came to a call with emails already on the canvas. Thomas and I spent the call discussing what he wanted to do next rather than giving a basic tour.
Cory walking Thomas and me through his work.
He became a paying customer after the call. That was more informative than an abstract persona exercise: a specific person, a real set of drafts, and a concrete next problem. The later Harmonya article separates our experience from Noah Adelstein's reporting on the broader program.
Thomas and I talk about polish in very practical terms. Can someone understand the next action? Does the preview preserve the design they just edited? Does the product remember the brand detail they already corrected?
Those details affect whether a new visitor gets through the first task. Attention from a launch gives you a chance to demonstrate them. It does not compensate for their absence.
For a founder planning a launch, I would make room for the follow-up work. Answer the questions, inspect where new users get stuck, and write down which customers find the product useful. A badge is a good moment. The work someone comes back to do is the better signal.
