Boxel Framework: How Cardstack Plans to Disrupt Application Development Using AI — Part 2

Understanding how Boxel and AI will work together

In our previous blog post on how Cardstack aims to streamline application development, we explored the complexities present in the current landscape and how the Boxel Framework will address them. This blog post focuses on how Cardstack is harnessing GPT and AI to empower developers and individuals to develop applications without the need to code everything.

To better understand this concept, let’s consider a typical workday scenario. We often rely on third-party tools to accomplish tasks, but these tools have a one-size-fits-all approach and lack the ability to be personalized to meet our specific needs. For example, with Google Forms or Hubspot Forms, customization options are limited, and users must choose from pre-set options or learn how to code to create forms based on their requirements.

Boxel changes this paradigm. With Boxel, users can not only address smaller use cases, such as creating forms using natural language but can also create complex workflows. Cardstack is integrating GPT with Boxel to enable users to create forms, invoices, and workflows using natural language. This is in addition to enabling developers to build applications using composable components.

In this blog, we will delve deeper into how Boxel will enable developers, businesses, and individuals alike to solve business-critical content management use cases and build applications using composable components.

How AI-enabled application development will work?

With the recent launches of generative AI-based tools such as ChatGPT and Bard, the tech industry is reconsidering its approach toward AI and how it can be leveraged to streamline current processes. We at Cardstack have been experimenting with it too and are now integrating GPT with Boxel. Our current R&D focuses on training AI to do what a developer would do, allowing AI to generate domain models for users at a fraction of the cost of a human programmer. To understand how AI will work with Boxel, let’s take a step back and understand how Boxel works.

[Note: It’s important to read our last blog in the Boxel series to understand many concepts from this blog.]

Boxel Framework focuses on the idea of building applications using composable components. The central idea behind this concept is to break down the process of building a domain model, which is a software stack that provides the necessary tools for creating different types of domain objects. The data for building a domain model using Boxel is encoded in JSON, and the schema leverages TypeScript, a superset of JavaScript that allows developers to create a web structure and reuse types for composing applications. The template is based on the web and uses Glimmer to generate HTML and CSS code. Furthermore, Business logic is implemented in JavaScript, enabling users to express everything in an Excel-like web application.

Boxels represent UI environments that enable users to connect different elements. The stack comprises several fundamental concepts, including boxels, cards, realms, and actions. Cards are a data model that employs fields within realms, allowing users to store and retrieve data from different sources. The system can also automate certain tasks using actions.

The system provides a uniform platform for developers to learn how to program in boxels and for users to use domain models. The AI can extend the system’s capabilities by generating more domain models through chat sessions or API sessions. Ultimately, the system aims to simplify generating objects and building applications using Boxels and AI, creating a composable application development ecosystem.

How Boxel’s composable components will be created?

We are working to integrate Boxel and AI to create a public library that will revolutionize content management. With the power of AI, Boxel will populate the library with thousands of reusable boxes that are already in existence. This catalog will function like an app store, but it is more of a template gallery where people can find pre-built components to use in their forms and workflows.

The library will allow users to save time and effort by not having to build everything from scratch. They can easily access the pre-built components and then assemble them based on their needs. The catalog will have basic components like an address, and custom components like medical software for dental offices, and other special cases.

Considering Boxel is an open-source framework, developers from outside can contribute to the library too. To contribute, developers will need to use the Boxel SDK and submit their work to create a comment. The library will be like generic drugs that are much cheaper than branded alternatives. Developers can contribute resources or get paid to build this catalog, and this will benefit users and also give developers a slice of the revenue generated from the use of their components.

What is exciting about this collaboration between Boxel and AI is that AI can generate simple software, and Boxel is designed to be simple too. With AI, instead of having thousands of developers working on one concept, one developer can use AI to create a tool that can generate all of the reusable basic common structures and interfaces that customers need. This developer can create a dictionary of basic concepts and then use AI to generate the Boxels for these concepts.

To make the catalog more useful, developers will need to train AI on how to generate Boxels in a specific way. They can do this by providing instructions on how to generate Boxels and what to generate. They can also use a context window to chain the different generative AI together and train the AI to generate Boxels from a dictionary of SAS.

The language of the description of the software is essential to the generation of software with AI. Once developers get the language of the description of the software right, they can send that information to AI to generate Boxels. The iterative process of generating the library is very much like data science, where developers try different algorithms and go against the data set to see what works.

The AI-generated software for all the Boxels in the catalog will be open-source. This means that anyone can use it and benefit from it. The open-source nature of the software will essentially make proprietary knowledge available to everyone. Different software companies create proprietary software, but with Boxel and AI, this knowledge can become part of the public domain.

How will non-technical users use Boxel?

Boxel is a powerful platform that enables businesses to extend and customize their workflows to meet their unique needs. By providing a library of common software components, Boxel allows users to build their own applications without the need for extensive coding expertise or technical resources.

One of the key advantages of Boxel is that it allows businesses to create applications that are both useful and adequate for their needs. Rather than settling for off-the-shelf solutions that may not be exactly what they require, businesses can use Boxel to build applications that meet their specific needs. By leveraging Boxel’s library, businesses can extend the functionality of their workflows to include specialized features that are tailored to their particular requirements.

For example, a shipping company may need to include special delivery instructions for certain addresses. By adopting the address field provided by Boxel, the company can customize the field to include a digital element for adding internal annotations within their workflow. This approach allows the company to extend the existing functionality provided by Boxel while maintaining compatibility with the general concept of an address field.

Boxel also enables businesses to customize the look and feel of their applications to align with their brand identity. By starting with an existing template, such as an invoice, businesses can adopt the template and then modify it to reflect their brand identity. For instance, they can add their company logo or change the color scheme to align with their brand. With Boxel, businesses can customize the application’s appearance without the need for extensive coding or technical resources.

Moreover, Boxel provides a constraint system called Guide, which allows businesses to generate formulas and enforce form fields to ensure data accuracy. This feature helps businesses maintain the integrity of their data by identifying and alerting users to missing or invalid data. This functionality can be expressed in code and generated by the GPT, making it easier for businesses to maintain the compatibility of their application.

Boxel’s real-time engine makes it easier for businesses to create and deploy their applications. Unlike traditional software deployment environments that require businesses to manage servers, hostnames, deployments, and databases, Boxel’s code lives on the platform. This approach eliminates the need for external services and enables businesses to use point-and-click tools to add or delete fields. This feature makes it easy for businesses to modify their applications without the need for extensive technical resources.

Boxel’s code editing tools can be replaced by chat sessions with the GPT, enabling businesses to make changes by describing their requirements in natural language rather than writing code or using point-and-click tools. This feature makes it easier for businesses to customize their applications, collaborate with team members, and share their applications with others.

How can users automate tasks using Boxel?

Boxel can be used to automate tasks and auto-fill forms using AI-generated data. Boxel provides users with the ability to create and customize forms and data schemas, allowing them to autofill information and automate their workflow using these forms.

One of the key benefits of Boxel is its ability to generate and transform data using AI. With Boxel and AI’s combined ability to generate code more efficiently and accurately than humans, users can take advantage of this capability to automate their workflow even further. For example, AI can fill out forms for users based on their data stored in Boxel, such as their addresses and seat preferences.

Boxel can also be used to automate risk management for financial transactions. By generating the parameters for an interaction, such as a blockchain instruction to rebalance a portfolio, users can rely on AI to make certain decisions based on predefined risk parameters. While it’s important for humans to approve these transactions, AI can take care of the heavy lifting, freeing up time and resources for the user.

One of the key benefits of Boxel is its open-source nature, allowing users to maintain control over their data and interact with it in a way that suits their needs. Our focus at Cardstack is to simplify software development by leveraging AI and open-source technologies. This is in contrast to some proprietary AI services that may use user data for their own benefit.

Want to learn more about how Boxel is leveraging AI? Watch our Founding Director, Chris Tse, discussing it here in detail:

About Cardstack

Founded in 2014, Cardstack is an open software ecosystem that enables Web3 projects to build and scale their business. The Cardstack ecosystem includes two product suites: Boxel Open-Source Composable Development Framework and Champer Revenue Management Protocol along with various tools for the Web3 ecosystem. With Cardstack, Web3 businesses can create and launch their own dApps, monetize their tools, and manage their revenue distribution. Learn more about Cardstack at https://cardstack.com