Q1: Learning to Build with AI
Zero revenue. Zero launches. Zero regrets.
I set myself a goal: build 30 small apps as a solo founder. No co-founder. No funding. Just me, AI tools, and hundreds of hours of research.
Q1 just ended. Retro threads trending everywhere. Most solo founders measure Q1 in revenue. I’m measuring it in research depth, competitor analyses completed, and ideas validated.
I have zero revenue. Zero launched products. But I also have four solid app ideas researched to the bone, a reusable tech stack figured out, and a process that I think can scale.
Here’s what honest Q1 transparency looks like when you’re at the starting line.
AI Hype
I am totally blown away by what I could potentially achieve with AI.
Honestly, it is absolutely amazing.
I am thrilled and tempted to start about a dozen of projects. Shaking hands prompting 20 sessions in parallel.
Yeah, I am that type who loves to kick off millions things but rarely finishes one. Once I learn that a solution exists I tend to move on. Some people call that a scattered brain.
I am old enough to know that there need to be some rules. And a discipline.
But before I get to those rules, let me tell you what keeps me up at night.
Are Developers Here to Stay?
Developers in our company already write over 90% of their code using AI. I’m not talking about autocomplete. I mean entire features, tests, refactors—all guided by AI agents.
That’s my company. Here’s another story.
A friend of mine runs a software development company. About 50 people. Last month he told me they stopped hiring junior developers entirely. Only seniors now. Not because juniors are bad. Because a senior with AI does the work of three juniors. The math is brutal.
But here’s what scares me. If nobody hires juniors today, where do the seniors come from in five years? Who learns the craft? We’re cutting the bottom of the ladder and hoping nobody notices.
I spent 20 years writing code. I’m good at it. And now I’m watching an AI write better code than most of my former teammates -- in seconds. That’s thrilling and terrifying at the same time. It’s the reason I’m building 30 apps. If AI replaces the *coding* part of my job, I want to own the *product* part.
Is the software developer a tribe that gets forgotten? Like the Frozen Empire:
Linus Torvalds suggests that engineers might still be needed:
I hope he’s right. But I’m not waiting to find out.
The Trenches
I need to dig trenches first. I would like to create a reusable template that I will easily copy for my apps in the future.
I also need to have a repeatable/scalable process.
For that I need to figure out a few things:
Web page, hosting, initial setup - done that.
Authentication - DB + social - done that.
Tech stack - Next.js, Tailwind, Shadcn... the usual suspects nowadays.
UX design - Penpot.
Payments - did that in the past in a very painful way. This needs rework from scratch.
Waiting list, Email and newsletter - this is probably the biggest time sink. Need to figure that under a single umbrella.
Localization - some indie hackers claim that this is useless.
Deployment - I need a cheap easy deployment. I no longer want to maintain Docker images myself. Likely Vercel + Supabase. Did that in the past.
AI chats - for apps that need it, LangChain and LangGraph, used that.
SEO - Claude SEO, Serpapi.
Did I forget anything?
I also want to be able to develop remotely, interact with my sessions from my mobile phone. Mac Mini price + energy consumption vs VPS. VPS is a clear winner. I’ll share my setup script later.
But let’s have some fun while learning first. I want to try out some things first.
My First Apps
I am a great fan of open-source, so I’ll probably start with an open-source app. Then I’ll do some SaaS because I am mostly experienced in building SaaS. And I always wanted to have my own mobile app in the AppStore.
My first apps already have names and competitor research behind them:
ClipJot (open-source) -- A clipboard image annotation tool. Grab a screenshot, draw arrows, add callouts, copy the result back to clipboard. I need this daily for communicating with my team and AI. If I need it, others probably do too.
SliderPress (freemium) -- A markdown-to-presentation converter. Write your slides in markdown, get a beautiful single-file HTML presentation out. I found multiple Reddit threads from people begging for exactly this. The existing tools are either too developer-focused or too clunky.
Links will follow soon.
Both have detailed competitor analyses and master plans done. That’s what hundreds of hours of Q1 research gets you.
I also need something for note-taking and todos. Sebastian Röhl writes about [building a “second brain”](https://sebastianroehl.substack.com/p/building-an-indie-app-business-88) -- a personal system that captures everything you learn so it compounds over time instead of getting lost. I want to build that for myself. A local-first note-taking app. Your data stays on your device. No cloud. No surveillance.
And I have an allergy tracker mobile app about 50% done from a while back. Maybe I resurrect that one for my first AppStore launch.
Best practices to start with AI coding
AI coding is not just Vibe Coding. The term author (Andrej Karpathy) himself now shifts to Agentic Engineering:
Let’s see what we can leverage here...
The Tool
I started with OpenCode first. The principle of plugging in any models looked like a perfect freedom. However, I ended up with Anthropic models anyway because they provided the best results.
I also ported/hacked in a couple of plugins from Claude.
And then one weekend I burnt down through all free credits on OpenCode Zen, OpenRouter, GitHub... And I spent over $100 with just one silly app (I started wrong anyway).
So a flat rate of Claude Max at $100/month is a bargain!
And when I switched, I also realized that Claude is better fine tuned for development. It seems to me to be a bit smarter, even when I used Opus and Sonnet on OpenCode, it was not as powerful. I would say 17.5% better :-)
Today, I saw an interesting video about the possibility to utilize Codex in Claude. So this might be a nice combo.
The Method
Then it is really simple. Read all the best practice documents. They have it all.
TL;DR?
Start with a very good CLAUDE.md! Keep it as brief as possible.
Keep sessions short, clear frequently.
Build, use, tweak skills for your specific tasks.
Use compound engineering or superpowers to span sub-agents.
And don’t forget to secure your agent:
What’s Next?
Q1 was research. Q2 is execution.
My plan for the next three months: ship ClipJot and SliderPress. Get them in front of real users. Write about every win and every failure here on Substack.
I have zero revenue and zero users. But I also have a clear process, a tested toolchain, and 30 ideas waiting in line. The starting line is behind me now.
I also have some letters of intent for custom business apps. Will see how that goes.
Let’s see what happens when I actually press “deploy.”