Diego Roshardt Grew AppAlchemy to 17,000 Dollars a Month in Four Months

An Idea That Started with Design Bills

What put AppAlchemy in Diego Roshardt's head was the cost of designing the mobile app he was building. He paid a designer on Upwork several hundred dollars, and small revisions brought extra charges. He thought that handling this process with AI could save other app builders time and money too.

He learned to code from YouTube during college and shipped several projects. Most failed, and the products that earned money sporadically never grew into a stable business. By graduation his friends had companies lined up, while he had no business results worth showing.

The products he had built still became a foothold for employment. He emailed startup founders to introduce his projects and was hired as a software engineer. A few months later he left the job to focus on his own product and moved into a studio apartment in Austin, with only a few months of runway.

The First Version, Where You Design by Chatting

At the time he was watching the spread of Lovable, which builds websites through conversation. He saw room to offer the same experience for mobile apps, where describing an idea produces a result. For people not used to installing tools and setting up a development environment, being able to start in a browser was an appealing proposition.

The first version of AppAlchemy came out in about two weeks. Because an earlier project had spent its time adding features while putting marketing off, he cut the development scope this time. He released only the core function first, where a user talks with the AI and previews the result.

What the early product solved was mobile app screen design. A user described the app they wanted to build, the AI generated screens, and they passed revisions through chat while refining the result. Finished designs could be exported as HTML or a share link. A prospective founder could hand the output to a developer, or show it to beta testers and investors, and move on to the next step. It was a way to see an idea in visible form before hiring a design freelancer or learning a design tool.

Diego charged a monthly subscription and a discounted annual subscription from launch. Paid customers arrived only sporadically at first, but those payments became the basis for judging demand. He did not set aside a period to gather free signups and then think about monetization.

Finding Customers on Reddit Without Followers

He had no large following to promote the product to. When Starter Story checked his X account while preparing its interview, it had about 70 followers. Diego found his first customers on Reddit, where potential customers were already gathered.

On Reddit he could approach readers by picking subreddits, the communities organized by interest. Using the community suggestions in the ad management screen, he built a list of boards related to his product, such as no-code. He researched where potential customers gathered without actually running ads.

His posts carried material that gave readers a reason to read, such as case studies of successful mobile apps. AppAlchemy was introduced as a tool connected to the story, and he compared the response to the case-study post and to the demo video. He described repeating this kind of posting two or three times a week.

In April 2025 he posted a demo to a community himself, called "idea to app design in 29 seconds." Along with the video he provided a link where the actual design could be inspected. Before reading a long feature description, potential customers could see what they would get. A video showing generation speed and an inspectable result stood in for the product pitch.

17,000 Dollars a Month in Four Months

In a Starter Story interview published in July 2025, he said he reached 17,000 dollars in monthly recurring revenue (MRR) about four months after starting. Signups won from Reddit were around 20,000, and product impressions passed one million. There was no paid marketing spend at the time.

In that interview, the largest cost was about 2,500 dollars a month for AI API usage. The development tool Cursor was 20 dollars a month, Vercel hosting about 40 dollars, and overall margin was about 70 percent, he said. AI costs incurred every time a screen was generated or revised were the business's main expense.

Billing was designed to keep generation volume manageable as well. Subscribers to the design product received a set amount of AI credits each month to spend on generating and revising screens. The structure collected a recurring fee while placing a limit on each customer's usage. It set the range of generation cost that had to be covered within the subscription a customer paid.

Swapping Tools and Expanding into an App Builder

In a July 2026 interview he also described changes to his operating tools. The early React product developed around Next.js, and authentication and the database moved from Firebase to Supabase. He used Cursor and Claude Code for development, choosing tools AI could handle easily.

In July 2026 he released a new builder that produces actual mobile apps. AppAlchemy 2.0 widened the scope toward generating React Native code from a user's description and checking and revising the running screen on a phone. The existing design product was kept as a separate service, so subscription revenue came in from both. The business's scope expanded from deciding how a screen looks to building an app that runs.

The new builder also combined a monthly subscription with credits. On the September 2026 price list, Pro was 20 dollars a month for 100 credits and Max was 200 dollars a month for 1,000 credits. Each message that built or revised the app deducted credits according to the workload, and users could export the generated code. The point was to start building inside the product while letting customers take the result with them.

Falling Revenue and a Sale

After the growth, a period of falling revenue arrived as well. On September 24, 2026, TrustMRR's payment-linked figures showed 240,073 dollars in cumulative revenue, 6,106 dollars in monthly recurring revenue, and 209 active subscriptions. The growth described in the 2025 interview had not continued unchanged.

In his sale listing, Diego said he had done almost no marketing since early 2026, and that this was the reason revenue fell. He also said he had gone through burnout and was looking at new projects, so he wanted to sell the business. Even while the product worked and subscription fees came in, someone was needed to keep acquiring customers and running operations. The developer who once gathered his first customers himself was now looking for someone to take over the product and its customers.