Meet the Agents at USV: Arthur, Ellie, Sally, and Friends

In the world of technology, innovation often thrives in unexpected places. A prime example of this is the story of Building 20 at MIT, a temporary wooden structure that became a beloved part of the campus for decades. Originally constructed during World War II to house the urgent and classified development of radar technology, the building was meant to be a temporary solution. However, it ended up lasting for 55 years and becoming one of the most productive buildings on campus.

The rapid construction of Building 20 was a marvel in itself. As described by Robert Kyhl in his article "Building 20: The Magical Incubator" (1998), the structure was built almost like those time-lapse photographs of skyscrapers, with workers efficiently collaborating to erect the building in real time. The flexibility of the space allowed researchers to shape it according to their needs, with the freedom to knock down walls and modify the building themselves.

Fast forward to the present day, and the concept of adaptable spaces has taken a new form in the realm of software. AI agents have made it possible for teams and individuals to customize their workflows, much like the researchers in Building 20. These agents, built around how specific teams operate, have enabled faster progress and increased productivity.

This Cambrian explosion of custom agents, tools, and software is not limited to engineering teams. Companies like Stripe, Ramp, and Coinbase have developed their own internal coding agents, such as Stripe’s Minions, Ramp’s Inspect, and Coinbase’s Claudbot. These tools are designed to harness the power of AI to streamline tasks and improve efficiency.

At Union Square Ventures (USV), we have been exploring the potential of creating our own internal agents and custom software to fit our unique workflow. Over the past three months, we have learned valuable lessons about the process of building these tools.

First and foremost, it is essential to start with solving one problem. When embarking on a project to create an agent or custom software, it is crucial to identify a specific challenge that needs addressing. This focus allows the team to develop a solution that is tailored to their needs and ensures that the project remains on track.

Once the problem has been identified, the next step is to gather a cross-functional team to work on the project. This team should include members from various disciplines, such as engineering, design, and operations, to ensure a comprehensive understanding of the problem and a well-rounded solution.

As the team begins to develop the agent or software, it is important to prioritize modularity and scalability. This means building the tool in a way that allows for easy expansion and adaptation as the team’s needs evolve. By designing with flexibility in mind, the solution becomes more robust and long-lasting.

Testing and iteration are also critical components of the process. As the team builds the agent or software, it is essential to conduct thorough testing to identify any issues or areas for improvement. This iterative process allows the team to refine the tool and ensure that it meets the desired objectives.

Finally, communication and collaboration are key to the success of the project. Regularly sharing updates and feedback within the team helps to keep everyone aligned and engaged. Additionally, involving stakeholders from other teams can provide valuable insights and ensure that the solution is truly meeting the needs of the organization.

In conclusion, the story of Building 20 at MIT serves as a powerful reminder of the value of adaptability and collaboration in achieving great outcomes. By applying these principles to the development of AI agents and custom software, teams can create tools that not only solve specific problems but also foster innovation and productivity. At USV, we are excited to continue exploring the possibilities of these agents and the custom software we are building, as we strive to create a more efficient and effective work environment for our teams.