Loading page…
Turn everyday ingredients and dietary preferences into recipes and meal plans, then monetize repeat use with consumer and business subscriptions.
about $900 MRR at dashboard snapshot · 97 paying subscribers
“It’s kind of like the succulent of software”: a small AI product that quietly earns while search keeps sending cooks to it.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Andrew Olsson launched DishGen, an AI recipe generator, about three years before the interview, after experimenting with an early language model and noticing that it could complete a recipe. He had run the food blog One Ingredient Chef, which gave the new product domain context and an existing site with links. DishGen lets guests generate recipes, then prompts them to create a free account for more use and saved history. Paid plans add near-unlimited generation, dietary and household profiles, longer chats, meal planning, recipe images, and commercial rights. A more expensive embed widget lets businesses put a customized recipe generator on their own site. Olsson described roughly $13,800 in sales over the prior year, about $900 in current MRR at the dashboard snapshot, and 97 paying subscribers. The business gets around 50,000 visitors monthly, with Google supplying most visits. It is a modest, low-maintenance asset rather than a fast-growing startup: prior audience, early timing, and search visibility did much of the distribution work.
For an AI recipe product, the first customers should come from people already searching for a meal solution, then from communities that share the same cooking problem.
The quick-reference card — acquisition channels, replication playbook, risks & traps, and our verdict — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself. The exact words are quoted below, with a timestamp that jumps to the second each figure is said.
“Over the last year, I did about $13,800 in sales volume.”
8:10 · hear it in the videoData credibility: Founder interview with Stripe and analytics dashboard figures described on screen; self-reported and unaudited. Google traffic and subscriber counts are also presented by the founder, not independently checked.
Make the first paid conversion happen at the moment a regular cook runs into a repeated constraint. A guest can test the recipe generator, then a free account can unlock saved history and more usage; when the user reaches a limit, offer a paid plan that removes friction and remembers diet, budget, and household size.
The compounding loop for DishGen is useful pages leading to more recipe use, more recipe history, and a wider set of indexed, specific recipe pages. Give every generated recipe a clean, stable URL only when it adds real value and passes quality checks; thin or near-duplicate pages can dilute a food site.
A founder building this kind of consumer subscription should expect quiet stretches where search traffic continues but revenue barely changes. Protect a small weekly maintenance budget, keep model costs and hosting visible, and choose a fixed review rhythm for support, safety reports, and failed generations.
DishGen wraps a general language model in a focused cooking workflow. Users can generate recipes from a natural-language request, ask for modifications, build a seven-day meal plan, save prior outputs, and create a recipe from a meal-plan slot. Personal profiles store diet, household size, and budget preferences so those constraints carry forward. Paid users can generate recipe images; business accounts can use commercial rights and generate images for distribution. A separate embeddable widget offers a branded recipe generator for other websites. The core model is a third-party model with custom prompts and product logic rather than a proprietary foundation model. That distinction matters: the defensible work is convenience, saved context, quality control, and distribution. The mobile app uses Flutter, but the founder said it requires extra authentication and deployment work and contributes only a couple hundred dollars monthly. Keep the web experience as the main product until mobile retention justifies that ongoing maintenance.
The main revenue source is subscriptions. A visitor can use the service as a guest, then register for expanded free usage and recipe history; reaching rate limits creates an upgrade moment. Premium plans add near-unlimited generation, personalization, and longer chats. A Pro or business tier gives commercial redistribution rights and more recipe-image generation, with the creator reporting 26 Pro subscribers and 69 individual Premium subscribers at the dashboard snapshot. One customer used the embed product, after DishGen previously had three or four. The founder described roughly $13,800 in sales over the prior year and around $900 MRR at the time shown, while 97 subscribers were paying. Annual plans create noisy monthly dashboards because the full payment is collected upfront. Mobile income is handled separately through RevenueCat, so platform-specific subscriptions should be reconciled with web billing before judging total revenue. The subscription model is understandable, though high churn and low conversion constrain its ceiling.
The web service uses a hosted application, a managed database, Stripe for web billing, and OpenAI models for text and image generation. The founder estimated a Vercel plan at about $20 monthly for this and other projects, an upgraded database at about $10 monthly, and model usage under $5 monthly at the snapshot. Those are direct operating figures from the interview, not a full accounting of labor, payment processing, taxes, app-store fees, or shared infrastructure. The Flutter app is connected to RevenueCat for iOS and Android subscription handling. A lapsed Apple developer renewal removed the app and interrupted app revenue until the account was restored, illustrating a low-cost but meaningful operational risk. For a new build, log model tokens and image calls per account, impose quotas, and monitor latency and failed generations. Keep generated content and profile data scoped to each account, and give users deletion controls. Low inference cost supports healthy gross margins, but churn and customer acquisition remain the harder economics.
Copy the narrow workflow: accept an everyday cooking constraint, return a useful recipe quickly, make revisions conversational, and save the result. Add persistent preferences only when they reduce repeated effort. The most valuable lesson is distribution fit: DishGen benefited from an established food blog and early search positioning, so a new founder should bring either a niche audience, domain expertise, or a credible search wedge before building a wide product. A modern copy should focus on a specific community such as budget meal planning, allergy-aware family cooking, or recipes for a particular appliance, then validate repeat usage with a small cohort. Avoid reproducing every feature at once. Start with guest generation, accounts and history, a usage limit, and one clearly differentiated paid tier. Defer a native app and an expensive embed sales motion until users ask for them repeatedly. Improve source citations, nutrition caveats, and recipe testing so the product earns trust beyond a polished interface. Measure retained subscribers, not just generated recipes or indexed pages.
The main risk is weak retention: the founder described monthly revenue churning by more than $1,300 over the prior year, close to the amount of new monthly revenue being added. Many cooks may need a recipe once or twice rather than pay continuously. General chatbots can absorb basic recipe generation, and competing wrappers can copy prompt workflows quickly. Search dependence adds exposure to ranking changes, especially if generated pages become repetitive or low quality. Model updates can change output quality or costs, while inaccurate ingredients or unsafe cooking instructions can damage trust. Consumer subscription pricing leaves little room for paid acquisition when conversion is low. A business widget may sound attractive but can require custom selling and support; this one had only a single customer at the time described. The mobile app adds a second product surface and renewal obligations for limited incremental revenue. Keep the scope tight, create genuine repeat-use features, protect search quality, and review store-account, billing, and safety operations regularly.
The four founder playbooks — first customers, first dollar, the flywheel, the silent stretch — are in the Unicorn Club, on this idea and on all 1,403 others. The story, the numbers, the sources and every later section stay open to everyone.
Join the Unicorn Club — $5/moCancel anytime, in one click