What is Vetrina?

Overview

Vetrina is a new user interface built from the ground up for large language models (LLMs).

It offers a complete workspace with tools to organize files, conversations and prompts into a personal digital knowledge base.

With Vetrina, you can take notes, make plans, study, write and review text, accompanied by the most powerful LLMs from top model providers.

Why?

Vetrina was created to solve deep user experience issues in AI chat interfaces.

As more powerful LLMs continue to be released every few months, the chat interfaces from all leading model providers (ChatGPT, Claude and Google Studio) still struggle with basic usability issues.

If you use ChatGPT daily, you can probably relate to the difficulty of doing basic things like:

1. Finding an old conversation from last week
2. Managing a complicated project with many different conversations and files
3. Losing important information because a conversation thread got too long
4. Uploading the same source documents repeatedly for each new conversation
5. Saving favorite prompts in a separate file or notes app by hand
6. Switching between conversations or keeping multiple browser tabs open

Despite these and other usability issues causing significant frustration, model providers seem more interested in shipping cutting-edge features that flex advances in LLM capabilities and tooling. This results in more bloat, visual clutter and unpredictability, and further compromises the essence of a conversational AI experience.

Our Vision

Vetrina is a conversational AI workspace that takes a new direction. Instead of chasing trendy capabilities, it aims to make using LLMs a delightful experience.

While we're excited by the periodic releases of ever more powerful models, as well as the growing ecosystem of LLM tools, protocols and frameworks, we believe there's still huge untapped potential in solving more fundamental challenges in how we interact with LLMs. To this end, we are interested in pursing the following directions:

1. Ergonomics

   As users and developers of power tools, we know that small conveniences add up fast. When a feature that usually takes several clicks and a few seconds is made instant, it becomes a mental "noop" and can be committed to muscle memory, freeing users to focus on higher-level work. A workspace built around these kinds of conveniences allows users to engage in powerful forms of creation and editing that wouldn't be possible otherwise.

2. Feature Composition

   We believe that a good workspace tool should be designed with a foundational user experience model that is simple, flexible and modular. This means that small individual features or capabilities are easy to understand and learn, and can be chained in new ways to create much larger and more powerful results.

3. Visual Clarity and Focus

   Tools for managing information should offer calm tidy interfaces that help users focus on their work without distractions. While it's tempting to try and surface all features by adding buttons and controls all over, screen space is prime visual and mental real estate that should be used thoughtfully.

These principles form the foundation of our vision. With them, we're aiming to build a beautiful conversational AI workspace that truly delivers on everything LLMs can do.