The Golden Age of Software Development: From "How Do I Code This?" to "What Should I Build?"
AI coding assistants are getting better. Their use is subsidised. The infrastructure costs of running applications in production are low. Now the question for indie developers is: what can we build?
The core challenge has shifted from "How do I code this?" to "What should I build?"
The barriers to creating software products are falling faster than ever. A convergence of AI assistance, cloud infrastructure development, and (current) industry subsidisation of AI and developer tools has reshaped what's possible for independent software creators. The one-person, billion-dollar company is well on its way! (It wasn't that long ago that we saw Instagram, 13 employees, sold for a billion. Now the goal is one person, one billion. Let's go!)
- AI tools are good and getting better.
- Their cost is subsidised right now.
- Infrastructure costs are also lower, and indie developers have options for low-cost ways to run applications in production.
I've written about indie developers before:
Curtis Collicutt
AI Enabled Software Development
Writing code is changing fundamentally. IDEs powered by generative AI can now effectively create code from scratch. Is it perfect code? No. Is it good enough? Probably. This includes the ability to create a design for a Software as an Application (SaaS) service, to build front-end and back-end components.
This shifts the core challenge from "How do I code this?" to "What should I build?"
The impact is significant: developers spend less time typing and struggling with the minutiae of development, although of course there are hundreds of tasks that AI cannot help with, and more time building something commercial. Overall, however, a task that once required time-consuming typing and connecting of sections of code can now be approached through the use of generative AI-based tools.
Using tools like Cursor's IDE seems like magic at first, for example using the composer mode where it writes files on its own. Over time it feels less like magic and can even be annoying to some developers, but in many situations using these tools is a boon to indie software developers.
AI Tool Subsidization
The industry's aggressive pricing isn't just about competition–it reflects a fundamental belief that AI processing costs will dramatically decrease over time, as we get better and better at running LLMs. This conviction is driving long-term strategy:
- Computing efficiency improves with each model generation
- Hardware specifically designed for AI inference is maturing
- Scale advantages have materialized, at least so far...
- Research breakthroughs are reducing computational needs
Infrastructure Costs Are Lowering
Independent of AI developments, cloud infrastructure costs have reached historic lows. (Although public clouds often have significant margins, especially on the virtual machines that underpin almost all platform abstractions).
Modern cloud platforms, such as Supabase, have removed the traditional complexity of running production software. It handles your entire database layer, from real-time updates to user authentication, while managing the underlying infrastructure, including backups. A developer can launch a full-featured application backend without touching a server configuration file. For many applications, Supabase's standard $25 per month tier provides more than enough capacity to support thousands of users. If a developer's application is financially successful, the $599 tier isn't out of reach.
But there's another route that's just as powerful and cheap. Companies like Hetzner have made raw server infrastructure surprisingly affordable. A powerful dedicated server with 64GB of RAM and multiple CPU cores costs around €40 a month. Abstractions like public cloud and Supabase have their advantages, but a simple virtual machine can support thousands of users for a SaaS application. This approach requires more technical knowledge to set up and maintain, but it is doable... and very cost effective.
Conclusion
The combination of these trends–the ubiquity and increasing quality of AI assistance, and the cheap commoditisation of infrastructure–is creating a uniquely favourable environment for independent software developers. The financial barriers to building and scaling software are disappearing, just as AI tools are making development more accessible.
For developers willing to build, the economic barriers that once made software development prohibitively expensive and time consuming are disappearing. The question isn't whether costs will continue to fall, but how quickly indie developers can take advantage of this shift.
Now the questions become more interesting. Like, "What software can I build that is useful to me?" And "What software can I build that can be financially successful?" Not just "How do I even build it???"