stats

Estimate output tokens saved by laconic this session.

Method

1. Scan assistant responses in this session.
2. For each response, count actual output tokens (≈ chars / 4 or words / 0.75).
3. Estimate "normal mode" length: multiply each laconic response by measured expansion factor 1.74 (from benchmark/results.md: 10172 / 5835).
4. Savings = estimated_normal - actual_laconic.

Output

Single line:

Saved ~N tokens across M replies (~X%). Bench factor 1.74.

If no laconic responses yet: No laconic replies this session.

Rules

• One line only. No table, no breakdown unless asked.
• Approximate — label as estimate.
• Skip non-laconic turns (e.g. before /laconic invoked, code blocks excluded).