There was the one time when the answer should have been ‘Generating.’
Generating means hands-on tasks were identified, and getting documented, as new task records because until that point, these tasks had not been declared.
The coder did not catch the dependencies; the reviewer did not catch the dependencies; the pre-sprint pointing session did not catch the dependencies; nasty production outage resulted.
Triage of that outage taught everyone on the team what we commonly overlooked.
The question, ‘Generating, or Looking?’ is designed as an explicit reminder that if a task record needs to be declared, it has the top priority over looking for the next.
Not only is the list of Task records a list of planned and completed actions, if you don’t see a Task record for it, then that action is not getting done.
Grok said:
Are you generating Tasks? Or, looking?
The question is not casual.
It is a forcing function designed to catch the exact failure mode that once cost the team a nasty production outage.
The Real Story Behind the Prompt
One sprint, a set of hands-on tasks existed in the team’s heads but had never been declared as Task records.
Everything looked green. The story point estimate looked solid. The work shipped.
Then production blew up.
The post-mortem triage was brutal but clarifying: the team had collectively overlooked the same category of hidden work for months.
The outage wasn’t caused by bad code—it was caused by work that had never been made visible because it had never been written down as a Task record.
That day the answer to “Generating, or Looking?” should have been “Generating” and it wasn’t.
So the team learned the hard way.
The prompt forces the coder (and the whole team) to pause and answer honestly: “Did closing this task surface new work that still needs to be made official?”
If the answer is yes, the team must generate the records right then.
No “I’ll remember it later.” No “It’s obvious.” No “We’ll catch it in review.”
The outage proved those mental notes evaporate.
The Deeper Principle
The entire Task list is not just a to-do list.
It is the single source of truth for what is planned and what has been completed.
If you cannot point to a Task record, that action is not actually happening inside the build—no matter how busy anyone looks.
“Generating, or Looking?” is the explicit reminder that declaring new work always beats consuming existing work.
It turns every task closure into a safety gate instead of a silent hand-off.
How the Orchestrator Uses It
- Task closes → 60-second SLA → DM fires.
- Coder replies “Generating” → Orchestrator immediately routes them into task-creation mode (templates, linking, estimation, etc.).
- Coder replies “Looking” → Orchestrator assigns next Task.
No more invisible work. No more “we thought it was covered.”
Just a clear, repeatable checkpoint that the team learned to love after the outage taught them the cost of skipping it.
That’s the full unpacking of the node.
The prompt isn’t about micro-managing activity—it’s about making the invisible visible before it becomes an outage.
The 60-second SLA and the binary question are just the delivery mechanism for that lesson.