Exclusive: Ex-Meta Engineers Unveil Jace, a Self-Sufficient AI Agent

Today, Zeta Labs, a London-based startup founded by former Meta engineers Fryderyk Wiatrowski and Peter Albert, announced the launch of Jace, an AI agent powered by large language models (LLMs) capable of executing in-browser actions on command.

The company also revealed it has secured $2.9 million in pre-seed funding, led by Daniel Gross, Y Combinator's former head of AI, and Nat Friedman, former CEO of GitHub.

Are You Ready for AI Agents?

With AI agents gaining attention—Cognition's Devin being one of the most talked-about—Zeta Labs positions Jace as a solution that doesn't require user guidance, enabling effortless task execution. Users simply need to instruct Jace on what to do, and it will handle the rest.

The startup is collaborating with early partners and plans to invest the pre-seed funds into enhancing Jace's reliability and speed for complex tasks demanded by consumers and businesses. Other contributors to this funding round included angel investors and VC firms such as Shawn Wang, Bartek Pucek, and ElevenLabs founder Mati Staniszewski.

What Tasks Can Jace Perform?

The idea for an AI agent emerged for Albert while he worked on an ecommerce project eight years ago, where he faced tedious operational tasks like data migration. With the maturity of language models, he joined forces with Wiatrowski to launch Zeta Labs and develop Jace.

Essentially, Jace functions as a web agent similar to ChatGPT. Users interact through a chatbox, providing task instructions in natural language or through widget-like prompts. Jace then creates a plan, gathers necessary information, and executes actions directly in the browser.

For example, if a user requests to book a hotel in Paris for a specified date, Jace will search the web for relevant information and autonomously book the hotel, including payment. Albert emphasizes that Jace enhances AI chatbots by embodying physical capabilities, enabling it to manage tasks ranging from simple flight searches and email replies to complex processes like building recruitment pipelines or launching ad campaigns. One remarkable instance even saw Jace create a company, complete with a business plan and client acquisition.

Autonomous Web Agent Technology

To achieve these advanced capabilities, Jace combines several models: a robust LLM for chat interactions that captures information and formulates plans, alongside Zeta Labs' proprietary AWA-1 (Autonomous Web Agent-1) model, which translates these plans into browser actions, effectively navigating the challenges of web interfaces.

Wiatrowski shared that the core model is based on open-source technology, enhanced through reinforcement learning from AI feedback (RLAIF) and extensive simulated interactions to prepare the model for multi-step web tasks. Jace's advanced cognitive architecture allows it to manage complex actions without looping back, accommodating hundreds of steps if necessary. It also includes security measures to encrypt user credentials for tasks such as LinkedIn job postings.

Future Prospects and Monetization

While Jace is capable of managing a variety of tasks, Zeta Labs has not yet monetized the product. The team is refining the AI agent with design partners in anticipation of a broader release. The next iteration of the AWA model aims to be larger and faster, enabling it to tackle longer and more intricate tasks, including those requiring visual interaction.

The majority of pre-seed funding will support this development and strategic hiring efforts. Zeta Labs envisions Jace as a valuable tool for consumers and small businesses looking to automate repetitive browser-based tasks across sectors like recruiting, ecommerce, marketing, and sales. A free plan will offer limited messages, after which users will be required to subscribe at a rate of $45 per month.

“We see significant demand from small businesses, particularly recruiters who manually source data from LinkedIn and transfer it to Airtable. This entire process can be automated with Jace; you simply need to ask,” Wiatrowski noted.

Most people like

Find AI tools in YBX