Teaching AI to code

80% of software code consists of repetitive tedious boilerplates. This code does not bring any business value but takes 20% of development time.

aicodoo uses AI to write boilerplate code automatically.

As there is no released product yet, this website gives you some insights.

How does it work?

For existing applications:

aicodoo learns coding according to your existing individual stack, integrations, GUI, and style. Then it writes scaffolding and boilerplate code for new artifacts.

For new applications:

Point aicodoo to an existing example application and it creates a new project that exactly reflects the example, but is adapted to your custom domain model.

Demo: AI-Generated Applications

Start the demo tool

Start the demo tool, enter a domain model, and AI will accordingly generate the source code of a new application.

Under the hood: AI has been trained on existing example code and composes new application code from the learned scaffolding and boilerplates.

Video: AI-Generated Code for Existing Applications

The following clip showcases an experimental IDE extension for adding artifacts to existing applications.

It's like an extremely powerful "code completion": Programmers just write very simple domain code and AI generates all the scaffolding and boilerplates from it.

The created code perfectly matches your existing architecture, programming stack, and coding style.

Under the hood

aicodoo is taught through Machine Learning on Source Code (MLonCode) to write scaffolding and boilerplate code.

The learned knowledge is used to write new code for already existing applications or to generate new programs.

Some AI technologies used by aicodoo: Code2Vec (Code to Vector Transformation), ML Algorithms, WEKA Machine Learning Toolkit, Apache Spark.

aicodoo learns and writes code by a given domain model. In the current status, only Entity-Relationship-Models are used.

aicodoo is language agnostic. It learns and writes any programming language and technology stack.

Current status: Research is done and AI technology is ready. Now trying to make some suitable tool out of it.

What do you think about it?

I am really interested in your feedback!
Feel free to share, discuss, and connect:

aicodoo on Twitter Alexander Derksen on Twitter Alexander Derksen on LinkedIn