← Previous: Using smart models for grilling · README
The practice
- Described as dead simple, yet so many people don't do it: grill multiple sessions in parallel (
10:25). - The author's loop (
10:25):- Answer the current session's question (usually by dictating, not typing).
- Switch to the second session, which has normally finished thinking by then.
- Answer its question, switch back — and bounce like that.
sequenceDiagram
participant You
participant S1 as Session 1
participant S2 as Session 2
You->>S1: answer question (dictated)
S1-->>S1: thinking…
You->>S2: answer question
S2-->>S2: thinking…
You->>S1: next answer
Note over You,S2: ~2× planning throughput
Answer to the "context switching" objection
- The author's rebuttal: it's really just managing two Slack threads at once — not that hard (
10:43). - You are making a lot of high-level decisions, but this is the only way he has found to increase throughput and get more planning done in less time (
10:43).
How many sessions
| Count | When |
|---|---|
| 2 | The normal maximum — his default limit (10:43) |
| 3 | Only when one session is on a long-running task such as research, and he's "feeling spicy" / high energy (10:43) |
| 4 | Suggested as reachable for someone with a more plastic brain, after mastering the basics (12:19) |