Replacing Micro-SaaS with LLM-Generated Code: A 20-Minute Experiment
Introduction
The idea that Large Language Models (LLMs) will disrupt the software-as-a-service (SaaS) industry has been a topic of discussion. While some claim that LLMs will make SaaS obsolete, others argue that SaaS provides more than just software, including compliance, correctness, and maintenance. In this article, we will explore the author's experience of replacing a $120/year micro-SaaS with LLM-generated code in just 20 minutes.
The Experiment
The author had been using a micro-SaaS called Shoutout.io to display testimonials on their website. However, when the billing section broke, the author decided to replace it with an LLM-generated solution. Using Codex, the author was able to recreate the testimonials section in just 20 minutes. The steps involved:
- Asking Codex to create a plan to remove the third-party dependency and host the testimonials in the codebase
- Tweaking the plan to use a modular approach with a separate JSON file for testimonials
- Adding a build step to generate the testimonials into HTML
- Testing and deploying the solution
Implications for SaaS and Software Engineers
The author's experience has implications for both SaaS vendors and software engineers. For software engineers, using LLMs to generate code can be a rapid and efficient way to replace third-party dependencies. However, it may require a certain level of comfort with the command line and verifying the output of the LLM. For SaaS vendors, the ability of customers to replace their services with LLM-generated code may pose a threat to their business model.
What This Means for SaaS Products
The author's experience suggests that rebuilding a SaaS is still a complex task, but rebuilding a specific use case can be relatively easy. SaaS vendors that do not provide ongoing value to their customers may be at risk of being replaced. Additionally, the ease of replacing SaaS with LLM-generated code may make buying and selling SaaS businesses less profitable.
Conclusion
In conclusion, the author's experiment demonstrates the potential of LLMs to disrupt the SaaS industry. While rebuilding a SaaS is still a complex task, rebuilding a specific use case can be relatively easy. SaaS vendors must provide ongoing value to their customers and maintain their products to remain competitive. As the use of LLMs becomes more widespread, we can expect to see more examples of customers replacing SaaS with LLM-generated code.