Task Spec And Structure
Front-load the task. Instead of revealing requirements progressively across turns, hand over intent, constraints, acceptance criteria, and file locations in the first prompt. Batch clarifying questions. Every turn adds reasoning overhead. For long-running safe work, use auto mode (Shift+Tab) so the model can iterate without frequent check-ins.
Effort Levels
xhigh is the recommended default for coding and agentic work.
| Level | When to use |
|---|---|
| Low | Cost- or latency-sensitive, tightly scoped work |
| Medium | Same as low, a bit more headroom |
| High | Balances intelligence and cost, good for concurrent sessions |
| xhigh | Default, best setting for most coding and agentic uses |
| Max | Genuinely hard problems only, overthinks on routine work |
Adaptive Thinking
Extended thinking with fixed budgets is gone. Opus 4.7 decides when thinking is worth it. Influence behavior with natural language:
- More thinking: "Think carefully and step-by-step before responding."
- Less thinking: "Prioritize responding quickly rather than thinking deeply."
Behavioral Changes
Response length now matches task complexity. State explicit style preferences in the prompt if you want something different. The model reasons more and uses tools less frequently. Subagent spawning is more judicious. For fanned-out work, explicitly ask for parallel agents.
Takeaways
- Stop dribbling context. Specs go in the first message.
xhighis the default. Don't bump tomaxout of habit.- If parallelism matters, ask for it by name.
- Control thinking depth with plain English, not dials.
