Agents hate friction: early thoughts on building for agents

Summary

Designing software for LLMs and agents requires more than extending human developer experience: a polished text-first interface can still produce an AX that is merely functional. The post frames an agent as a combination of model, harness, context, and skills, with behavior varying across providers, models, and harnesses; it focuses attention on commonly used combinations while testing beyond them. In a test setup, agents receive goals and different amounts of context, and model-harness combinations are evaluated by how they reason, which tools they call, how many steps they take, and how consistently they repeat a task. The resulting design questions include friction, context and token efficiency, familiarity with established tools, and the trade-off between fewer steps and higher context or cost; the central conclusion is that agents may abandon a product when friction persists, making testing necessary to expose that behavior.

Takeaways

1. 1

   An agent’s behavior depends on more than its model: harnesses, injected context, and optional skills also shape how it uses a product.

2. 2

   Repeated testing can measure variation between agent runs, including tool selection and call counts, and can reveal whether design choices improve consistency.

3. 3

   Reducing friction involves trade-offs among context size, number of steps, token cost, and familiarity with established tool patterns.