(agents)
A new layer of software is emerging - agents that can understand requests, make decisions, and take actions on our behalf. These systems can interpret context, adapt to new situations, and execute complex tasks.
This ability requires careful design. When software can act independently, we need to consider capability, constraint, transparency and control.
OpenAI and Cursor are good examples. Operator has a fully interactive browser that shows the step-by-step decisions of the agent, while Cursor takes a productized approach: Chat never makes changes without approval, whereas Composer is an active coding partner with some autonomy.
In the AGI race, the design challenge is to build believable, trusted software experiences.