My coding agent is working. Should I wait—or start something else? 🤔
That’s one of the more unexpected challenges of agentic coding.
If the agent needs another twenty minutes, I could tackle a different topic. If it needs a decision in thirty seconds, switching context might cost more than it saves.
And sometimes, just as I get into that second topic, the first agent comes back with a question.
What I want to know isn’t only:
“When will the code be finished?”
It’s:
“When will this task need my attention again?”
I’ve written two articles exploring that problem:
💡 The story: time planning, context switching, and the awkward gap between delegating work and reviewing the result.
https://berndkuhlen.de/notes/ai-coding-assistant-honest-eta.html
🛠️ The practical companion: a small Python example for reporting attention states and rough time windows—and clearly marking estimates that have expired.
https://berndkuhlen.de/notes/agent-eta-attention-window-howto.html
The example uses explicitly reported states and estimates. It doesn’t pretend to predict an agent’s progress automatically.
How do you handle the gap: stay with the agent, switch tasks, or run another agent?
Top comments (0)