Building with AI: The Solo Founder Era

·Val Kamenski·5 min read·#Startup #AI

When I started my first company in 2012, turning software into a real business usually meant building a team first. At a minimum, you needed two developers and a part-time designer, and enough savings to survive a few months while building and validating your startup.

For most people, that was enough to kill the idea of a tech business before it even started: too much risk, too much money, too many unknowns.

But that's all changed, first with remote work going mainstream, and now with AI. Welcome to the solo founder era. The age of the solopreneur.

It's never been this easy and this fast for a single person, even without a technical background, to build a website or mobile app in a few weekends. It's never been this cheap to publish it for the entire world. No more burning through thousands of dollars just to get started.

And I think it's only going to get easier and cheaper.

What has actually changed?

1. AI now generates more reliable code than two years ago

Two years ago, founders asked me: "How do I find a technical co-founder?" Now they ask more: "How much can I build myself?"

With Claude, Cursor, and Codex, a non-technical founder can build a working prototype. You still need engineers for complex systems, but "complex" means something different now.

Definitely don’t go live without a thorough code review, especially for anything high-stakes that could harm people. But for showing a working prototype to early adopters or interested users, one person can handle that now, even without deep technical knowledge.

And it's not just code. Services like Midjourney, fal.ai, or ChatGPT can generate professional visuals in minutes. Need a voiceover or sound effects? ElevenLabs creates authentic audio that used to require a recording studio. Need a demo video? Runway can help with that too.

A solo founder can now handle far more than before.

2. Infrastructure became more affordable, or even free

In 2012, setting up servers meant wrestling with Linux configs, load balancers, deployment pipelines, and a bunch of other voodoo. Today, you can ask Claude Code or other models to guide you through much of it. Plus, many of the services below let you deploy your project through a web UI. Most also offer free tiers for people just getting started.

For example:

For most early-stage products , "DevOps person" optional now. And you don’t need to build everything from scratch, like payments, because you can ask AI to help integrate your project with existing services.

3. Remote work became the default

Back then, remote work was unusual. You had to justify not being in an office.

Now it's the default. Which means:

  • You can hire the best person, not the closest person
  • Your costs can be dramatically lower
  • Time zones become a feature ("someone's always working")

What AI can't replace?

Before you think I'm saying "AI solves everything" - I'm not.

After 13 years of building, here's what I know still requires humans:

Understanding what to build. AI can write code. It cannot tell you if anyone wants what you're building. The hardest part of startups was never the code. It was figuring out what problem actually matters.

Selling. No AI is going to hop on a call with your first ten customers and figure out why they're not buying. That's still you.

Persistence through the dark moments. Every startup has a moment (usually many) where everything feels broken. AI won't push through that. You will.

Taste. Knowing what "good" looks like. Knowing when something is 80% done versus truly finished. This is still deeply human.

If I were starting over today

Sometimes I think about what I would do differently if I were starting from zero in 2026.

I wouldn't try to build a big team fast. I'd stay small as long as possible.

I'd use AI aggressively for everything it's good at: first drafts, code generation, research synthesis.

I'd stop obsessing over perfect code. How often does code actually stay the same? Competitors ship features, libraries update, requirements change. You rewrite things anyway. After a few months, you forget what half the code even does. And let's be honest, your code isn't perfect either. AI generates decent, working code now, so use it.

After all, customers don't pay for ideal code. They pay for a product that solves their problems.

I'd build only what's absolutely necessary to test my idea. Every feature you add before finding users is a guess. Most guesses are wrong. Ship the smallest thing that could work, then spend the rest of your time figuring out how to get it in front of people who might actually pay for it.

And I'd move faster than I think is comfortable. Because the cost of trying things has dropped so dramatically that the main risk isn't failure - it's moving too slowly while someone else figures it out first.

The bottom line

If you've been thinking about starting something, the barriers have never been lower. Building a business is still hard. Finding customers is still hard. But starting? That part just got easy.

This is the second in a series of posts about what I wish I'd known when I started. The first one, "From Fear to Founding," covers the emotional side of taking the leap. If you're an early-stage founder working through these decisions, I offer free mentoring sessions no pitch, just a conversation about whatever challenge you're facing.