First, thank you for participating in our beta. Your participation will help make this product the best it can be. This article will address the guiding principles around Patentext and provide a basic outline of our UX.

To understand the design choices behind Patentext, it’s helpful to consider the status quo. Whenever a prompt is submitted, other AI patent drafting tools use Retrieval Augmented Generation (RAG) or a similar method to access relevant material across the many documents you contribute to these applications. These documents can include:

• invention disclosure forms
• a transcript of a call with the inventor
• a draft set of claims
• an outline of the specification
• a problem-solution statement
• your conversation with their AI assistant
• the patent application in progress
• And more

With so many inconsistent documents to draw from, the response of the AI is difficult to predict and often inaccurate. To address this, these tools rely on many subsequent conversational prompts from the user to refine the output.

How other AI drafting tools work:

Patentext differentiates itself from other AI drafting tools by opening up this black box of AI context. Patentext takes the initial disclosure materials listed above and creates an abstraction of the invention just as a patent practitioner would normally do. We call this abstraction the “Invention Graph,” and it functions as the source of truth for all prompts you create in Patentext.

How Patentext works:

Patentext leverages the invention graph to provide many helpful features such as our automatically generated outlines, Section Blueprints, and Node References (coming soon), which are described in further detail in other articles. Most importantly, the invention graph ensures that all prompts submitted to Patentext draw on accurate, practitioner-approved content, greatly improving the technical depth and accuracy of our generated text.