3 min read

The micro-SaaS challenge: Day 80 (Building Apocode #10)

The micro-SaaS challenge: Day 80 (Building Apocode #10)
Photo by Markus Winkler / Unsplash

I wrote previously of my intention to create and launch 10 micro-SaaS products in 100 days, while building Apocode (a UK-based startup helping entrepreneurs create SaaS products without writing code).

Ten days ago, I completed the seventh micro-SaaS. This article introduces the eighth.

The rise of AI

The concept of artificial intelligence (AI) is an old one. The ancient Greeks invented the legend of Talos, a 30m tall bronze robot designed to protect the island of Crete from invaders. In the 11th century, the Mesopotamian scientist Al Jazari built the first recorded programmable automaton. And in 1760, the French watchmaker Pierre Jaquet Droz created the first programmable computer.

The Writer, the first programmable computer created around 1760

AI development accelerated with the advent of modern computing. As an example, game AI was barely capable of playing a game of chess or checkers in the 1950s. But by 1997, the chess AI: Deep Blue was able to defeat the world chess champion, Gary Kasparov.

click to embiggen
The first match between Gary Kasparov and Deep Blue in 1996

Today, AI is everywhere - and keeps getting smarter. Smartphones are unlocked through face recognition, voice assistants facilitate customer support, and cars can drive themselves.

For a long time, AI was limited by computing power and the data available for training. But these problems are resolving; today's computers are multiple orders of magnitude faster than their 1950s equivalents. And we live in the age of big data, where tech companies process and store petabytes of data per day.

Somewhere in the distant future, humanity might even reach the point where they are able to create self-aware AI, equipped with the ability not only to perform any task better than a human being, but also capable of self-improvement. When/if that day comes, it will be either an amazing force for good or the downfall of civilisation as we know it.

Blog post generation

For my eighth micro-SaaS, I decided to explore the possibility of creating a blog post generator powered by AI.

Building such a model from scratch would be extremely difficult. So, instead, I chose to use one of OpenAI's text generation engines, GPT-3. It was made available for public use by OpenAI in 2020, and a major upgrade was released earlier this week. It is supposed to be much better at long-form text generation, so a blog post generator is a good way to test its capabilities.

The resulting micro-SaaS, AI Blog Creator, is very simple. It asks the user for their intended blog post title or idea, which it uses to create a list of subtitles through GPT-3, and then uses GPT-3 again to fill each section with content.

Does it work?

The quality of blog posts created by the AI varies widely based on the topic or question entered.

For example, when asked "Can AI really create a blog post?", the AI produces a relatively coherent blog post:

When given other topics, the results are mixed. For example, I provided the AI with the title of one of my previous blog posts to see what it could come up with. Here is the result:

Overall, the AI created a lot of interesting material, but it misinterpreted certain titles/prompts which resulted in some irrelevant content. All generated posts would still need to be carefully reviewed by a human before publication.

You can test the blog post generation here: AI Blog Creator.

If you would like to provide feedback, create similar SaaS products, or follow the journey, you can join Apocode's product waitlist, subscribe to my blog, follow me on Twitter, or email me at nicolas@apocode.com