ThirdLayer

ThirdLayer is an applied research and product lab seeking to understand the learning policy behind personal productivity agents. Specifically, in an age dominated by information and knowledge work, we believe that software is most productive for a user when it reflects each person’s unique workflows, ideas, and outputs; any app that you use should require minimal onboarding, and it should proactively meet you wherever you want to use it.

The context management bottleneck

Today's most powerful models have incredible capabilities across a number of fields: achieving gold-level performance on the IMO, perfectly solving all 12 2025 ICPC problems, and giving scientists the ability to predict protein structure. However, these capabilities are minimally accessible in a real-world environment because models remain stuck behind static chatbot sites and have no access to the goal-specific context that you’re using in other apps or platforms. Furthermore, because of this lack of proactivity and contextual understanding, using AI platforms is itself a context switch, which introduces significant friction if you want to leverage these tools to work on your task.

On-preference interfacing between humans and agents

To break down these barriers between humans and their tools, we’re building Dex, a browser extension-based agent that continually updates its behaviours based on each action you take in your browser. By operating within your workspace and learning directly from real-world, first-party task execution, Dex is able to expand the surface area from which it collects your user intent, allowing it to stay on-preference within your unique workspace and pattern-match itself to you.

Join us.

Our team comes from top universities and frontier labs across the world. We are resarchers and engineers who have worked at the top quantitative resaerch firms, academic labs, and frontier labs like Jump Trading, Point72, Meta. Furthermore, we are fortunate to be backed by advisors and collaborators in research and design from MIT Media Lab, Stanford SAIL, Berkeley BAIR, Sakana AI, DeepMind, and Eleven Labs.

We're pushing forward the product frontier through our technical expertise and research and hoping to create an artificially intelligent product whose benefit is not constrained by the bandwidth of the team behind it.