so much of AI Engineering isn’t an intelligence play; it’s an architecture play. Even many of the older models have a lot of untapped intelligence that can be unlocked by good design

• Devansh (Artificial Intelligence Made Simple)

This feels right to me. All of these agents (ie. Manus, Deep research) feel like they are as much or more about good engineering as they are about the models they use. A very specific/clear task + smaller/older model + tools = a seemingly brilliant response.