Tarx is Smarter Than ChatGPT: Delivering on Our Promise

Over a year ago, in order to trick Tim into working on the Tarx, I told him I thought I had an idea for an AI thingy that could be smarter and better for humans than the LLMs the big guys had launched. After he laughed, he did the research, he tested, and we started working with partners like Nvidia, Baseten, and Nebuly, and ran alongside some great open-source projects like Sycamore, LangChain, and well - Tim started to believe it. We rounded up some really smart people and set out to build Tarx with the promise that it would surpass ChatGPT in intelligence, productivity, and increasing human capacity, all while doing it ethically and not stealing human insights, intuition, context, and know-how.

As we approach our launch, we’re proud to say that Tarx delivers on that promise—and more. By combining 1000+ open-source AI models, real-time data connections, robust ethical safeguards, and dynamic ai-agent creation, all wrapped in the first ever Human Language Model (HLM), Tarx is the fulfillment of a vision of a better way humans and AI can collaborate in the workplace and beyond.

Recent studies, such as the one conducted by Harvard Business School, have shown how tools like ChatGPT significantly improve productivity—boosting performance by up to 40% in areas like writing, analysis, and problem-solving (1), (2) . These findings underscore ChatGPT’s ability to answer questions and create work product, setting a high standard for AI-human collaboration. However, we designed Tarx to augment human abilities.

While tools like ChatGPT and Anthropic’s Claude are incredible at handling text-based generation and language tasks, Tarx takes a more 'integrated into your daily life' approach. It learns you, learns how what your job is, what's hard for you, what's easy, and what are the thinsg you care about. We designed Tarx to dynamically adapts to specific job roles based on who you are. Using a multi-model architecture, Tarx pulls from thousands of specialized AI models, each optimized for particular functions. This ensures that each task—whether it’s writing, analysis, or communication—receives the most effective AI model, significantly increasing performance across the board.

Benchmarking & KPIs:

To ensure Tarx surpasses the productivity gains seen with ChatGPT, we focus on several key metrics:

• Task Completion Time: Tarx's ability to dynamically select models specific to the user’s job role results in time savings of 10-15% compared to generalized AI tools .

• Accuracy and Quality: By utilizing LangChain for print management and task-specific integrations, Tarx improves the quality of outputs by 5-15% .

• Adaptation Time: Tarx’s Human Language Model (HLM) learns and adapts to users faster, reducing the AI’s learning curve and allowing for a 15% faster onboarding time for new users.

2. Ethics and Human-Centered AI

At the heart of Tarx is a commitment to ensuring that AI remains on the side of humans. We prioritize ethics, making sure that:

• Human insights and data are securely stored and only used by the HLM for the specific user. Tarx never shares or processes user data outside its intended purpose, keeping human interactions private and secure.

• Tarx focuses on empowering users rather than replacing them. This is reflected in our AI design philosophy, where Tarx augments human abilities by working alongside people, not as a replacement for their creativity or decision-making.

Talk to Tarx: Advanced Model Integration: Vision, Speech, and Text

Tarx combines OCR (Optical Character Recognition) with leading text-to-speech models, enabling it to see, speak, and chat with users in real-time. This creates a multimodal interaction experience that’s unmatched by traditional language models like ChatGPT. Tarx’s ability to see what you see, capture information from documents, and respond verbally or through chat offers a complete, integrated experience that fits seamlessly into everyday workflows.

Capabilities in Action:

• OCR & Vision: Tarx uses OCR to process documents, images, and even physical screens, extracting data and offering contextualized suggestions or actions .

• Speech Interaction: Leading text-to-speech models allow Tarx to not only understand but also communicate verbally, enhancing its usefulness in meetings, brainstorming sessions, or hands-free environments.

• Chat Interface: Backed by a conversational chat interface, Tarx can provide written insights, analysis, and recommendations based on the user’s needs.

4. Sycamore and LangChain: Seamless Data Management

Tarx integrates both Sycamore and LangChain to handle unstructured data and facilitate smooth print management and workflow integration. Sycamore acts as an LLM-powered search and analytics platform for handling large volumes of unstructured information. LangChain helps Tarx organize and manage workflows, particularly in environments where handling print and document-based tasks is critical. Together, they enable Tarx to partition, search, and organize data more effectively, providing users with quick, actionable insights .

Why Sycamore and LangChain Matter:

With Sycamore, Tarx can effortlessly handle unstructured data like reports, documents, and emails, making it ideal for professionals who deal with high volumes of information. Combined with LangChain, Tarx ensures that workflows are streamlined, from content creation to print management, enabling the AI to manage tasks at a deeper, more integrated level.

5. Dynamic Models: Specialized AI for Specific Tasks

Tarx has access to thousands of AI models, which it dynamically pulls together to ensure higher fidelity and higher-quality output compared to ChatGPT. Tarx doesn’t rely on a single model but instead contextually selects and combines models based on the task at hand. This flexibility results in more accurate and efficient task execution tailored specifically to the user’s role.

Just a few of the open-source models in use:

• Meta-LLaMA 3.1: Tarx leverages Meta’s latest language model for high-quality instruction-based tasks, such as content generation and report writing.

• CohereForAI’s c4ai-command-r-plus: Designed for conversations and tool use, ensuring Tarx excels in real-time collaboration and decision support.

• Mistral Mixtral 8x7B: Ideal for expert-level tasks requiring precision, such as data analysis or complex project tracking.

• 01-ai/Yi-1.5-34B-Chat: A powerful multilingual model with text and image conversation capabilities, enhancing Tarx’s ability to support international teams and visual data processing.

With that said, Tarx is More.

Tarx isn’t just another AI tool; it’s a tool for humans. It’s a sidekick that adapts, learns, and grows with you. By focusing on deep integration with human workflows—from seeing what you see to talking back with insights and accessing your favorite apps—Tarx delivers on the promise to be smarter than ChatGPT. Our benchmarking and KPIs prove Tarx’s superior ability to save time, improve quality, and empower users to do more with AI at their side.

With Tarx, the future of work is not just automated—it’s augmented.

-John

Written with Tarx

This article was created with a blend of inputs—approximately 70% directed by John, including strategic insights, content structure, and conceptual direction. The remaining 30% was augmented by Tarx, assisting with research gathering, content enhancement, and refinement of technical details.

Some of the capabilities outlined here, including deeper integration with specific AI models and workflow automation features, represent forward-looking statements about Tarx’s evolving functionality and vision for future releases. To access new bells and whistles, you'll need to request access to beta rollouts.