Claude Code’s Next Era — Thariq Shihipar, Anthropic
Source
latent.space
Date
Context
Useful prompting depends on knowing what information the AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → needs. In this interview, Anthropic'sThariq Shihipar describes it as learning both how the model behaves and the vocabulary of the task. His design example: someone who is not a designer might ask for eight mockups, whereas a designer could name reference sites, a typeface, and specific components. Those details reduce the choices the model must guess. He also argues that spending more effort up front on context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → can avoid cycles of asking the agent to undo and redo work, which also consume usage. This is advice from his experience with Claude Code users, rather than a controlled productivity result.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
