Back arrow
Go back to all blog posts
MutableAI, AI Coding Assistant for Jupyter

MutableAI, AI Coding Assistant for Jupyter

What is MutableAI?

MutableAI is a new kind of coding assistant that will make you more productive by automating the tedious bits of your life as a data scientist. The main goal of the extension is to save you time and get you to production faster.

NLP research has taken flight over the past couple of years resulting in breakthrough achievements. Now, a new wave of NLP based coding assistants are being developed that take advantage of these results. By pre-training on vast bodies of publicly available code, these intelligent models are able to recognize patterns that can serve as low-level building blocks for intelligent code automation tools like MutableAI.

Today, there are multiple tools on offer. Alternatives to MutableAI include GitHub's Copilot, Amazon's CodeWhisperer, Tabnine and Kite. Today we're announcing that we're partnering with MutableAI. We were deeply impressed by the founding team's vision & execution. Furthermore, their focus on first class Jupyter Notebook integration meant we could make MutableAI available to Orchest users in just a matter of weeks!

Features of MutableAI

These are the main features of MutableAI today:

  • Autocomplete
  • Open ended transforms
  • Productionize code
  • Type annotations

Autocomplete

Usually, when coding, you're not the first person to do something. Even if your details vary slightly, the general structure of the operation you're performing has probably been performed many times before.

Why then, do we bother to type everything out manually every time we: interact with an API, unwrap a result object, or transform from one data structure to another?

The time has finally come where tools can take advantage of all the existing code that has been written to make repeating yourself a thing of the past. By intelligently recognizing the underlying patterns and slotting in your specific context (variable names, file structure, scope) the machines can now generate an accurate auto-completion that gets you to your end result in significantly less time.

Open ended transforms

Sometimes you wish you could just ask someone to make the relevant changes based on a high level instruction. You just wish someone else could "figure out the details". Especially as you become more experienced, dealing with the specific details of the "how" can get boring pretty quickly.

MutableAI offers domain specific transforms that understand your code seamlessly. Simply use natural language to describe the desired change:

The result: šŸ¤Æ

Productionize notebooks

When writing an initial version of your notebook, you're likely iterating quickly and adding code incrementally.

Once you're satisfied with your result, you'll likely want to tidy up to increase code readability. For example, you might group all imports at the top to follow import conventions and to make it easy to see at a glance which modules this notebook depends on.

Type annotations

Many modern programming languages support incremental type annotation, or said more simply: adding some types as you go. As a human, we can often infer which types a function expects based on the body of the function. Can't a machine analyze code and add types in a dynamic language context in the same way humans would? Due to recent advances in pattern recognition and NLP techniques, this is now feasible & usable at scale.

Future of NLP coding assistants

One of the most exciting features of MutableAI is the open ended query dialog. It sparks your imagination and opens up the possibility of ever more ambitious and complex queries. Imagine a future where you could give high level instructions to your IDE like "replace pandas for polars" or queries like "improve iteration performance by using more caching".

With NLP advances following one after another we're certain these tools are just scratching the surface of what's possible. Human brain cycles are expensive and we can't wait for a future where more tedious work is delegated to machines.

What are you most excited about? Which Copilot or MutableAI moment has surprised you the most? Share your thoughts in the comments below or on social.

Want to try it out yourself?

MutableAI is now available for free in Orchest Cloud

Create a free tier Orchest instance at https://cloud.orchest.io/signup

What is Orchest?

Orchest is an open source tool that lets you easily create batch data pipelines. Find out more on our homepage.

Orchest pipeline editor

Opt-in to the MutableAI extension and start experimenting!

MutableAIĀ in the integrated JupyterLab editor