In the wake of GrowthHax we've continued to hone our suite of products. This means not just fixing bugs in Nolita, but making decisions about the future of the suite. Autonomous web navigation is still an active research area for many companies, and we have a very idiosyncratic approach, selectively (and agnostically) deploying models in the automation pipeline. We want to rely on the strengths of our approach, but stay open to redesigning other aspects of our approach that still require more careful planning.
Some of these changes require some heads-down work, and we anticipate to be in an exploratory mode as a company for the next quarter in addition to our regular maintenance cadence.
In the short term
We're currently incorporating local models into Nolita as a provider, specifically Ollama as a starting point. We also have begun work incorporating explicit planning into the agentic framework before beginning the navigation process.
Looking ahead
With this work as a first step, our goal is to then reconstruct the way we prompt the Memory Index with planning in mind; we currently rely more on being clever with searches and queries for replaying navigation steps, but we plan to move toward evaluating graph similarity for identifying similar pages.
In addition, we will continue to complement our suite of applications that extend and connect new intelligence. We're going to begin incubating additional projects that supplement and compliment Nolita.
We will post updates we have more to share about our efforts. As always, we are grateful for those of you following our work.