How one developer built two profitable AI apps to $100K ARR on the side

Executive overview

Most developers assume building a profitable SaaS requires quitting their job. Ramsri runs two AI apps generating $8,500/month combined while working full-time and spending just two hours a week on them.

The key is choosing ideas that automate manual tasks people already pay for, building distribution before the product, and treating the apps as autopilot businesses once launched.

Consistency in small daily actions compounds into a business that runs without you.

The two apps and their numbers

• Supermeme.ai: $130K lifetime revenue, 1.5M+ users, 400K paying customers
• Question.ai: $100K lifetime revenue, 150K+ users, 182K–200K paying customers
• Combined: ~$8,500/month, $100K+ ARR
• Profit margins: 70–75%; biggest cost is API usage (OpenAI, Anthropic)
• No employees beyond the founders

How Ramsri manages time

• Keeps full-time job deliberately — avoids VC funding, hiring, and team management
• Apps run on autopilot; dedicates one Saturday afternoon (2–4 hours) per week
• Initial phase required 1–2 hours of daily learning across tech, payments, auth, SEO, marketing
• AI coding tools (Windsurf, Cursor) now make building and maintaining significantly faster

Idea-finding framework

• Target manual tasks people currently pay for that AI can significantly enhance or replace
• Identify the use case before the technology is ready — anticipate where generative AI is headed
• Question.ai insight: large edtech and publishing companies paid for manual quiz creation; AI could automate it
• Validate by tracking AI capability curves, not just current demand

Building distribution before the product

• Write publicly about what you're working on before launch — blogs, LinkedIn, YouTube
• Consistency matters more than individual post performance (10 likes or 400 likes, keep showing)
• Building in public attracted early users, champions, and even co-founders
• Launched on Product Hunt with an existing audience; Supermeme hit #1 on launch day
• Virality of the AI meme format amplified initial growth

Tech stack and build approach

• Frontend/backend: Next.js (full-stack JavaScript)
• Database: Supabase
• Landing pages: Typedream (no-code, rapid SEO iteration)
• AI coding: Windsurf, Cursor
• First version of Question.ai was no-code: two text boxes, built quickly to test the idea
• Rewrote on modern full-stack after validating the concept

Advice for early-stage builders

• Think like a builder from day one — viable prototypes can be built with Lovable, Bolt.new, Make, n8n, Zapier
• Aim to chain and stitch existing tools rather than spending months in R&D
• Follow AI capability updates closely to spot new use cases before they're obvious
• Build a minimum viable prototype for any promising idea immediately