{% extends "base.html" %} {% block title %}aide - Insights{% endblock %} {% block content %}
How first prompt length correlates with session outcomes.
| Prompt Length | Sessions | {% if not subscription_user %}Avg Cost | {% endif %}Avg Errors | Avg Edits |
|---|---|---|---|---|
| {{ b.label }} | {{ b.n }} | {% if not subscription_user %}{{ "${:,.2f}".format(b.avg_cost) }} | {% endif %}{{ "{:.1f}".format(b.avg_errors) }} | {{ "{:.1f}".format(b.avg_edits) }} |
How cost is distributed across sessions — a few whale sessions often drive most of the spend.
| Session | Project | Cost | Cumulative |
|---|---|---|---|
| {{ s.custom_title or s.started_at[:10] }} | {{ s.project_name }} | {{ "${:,.2f}".format(s.cost) }} | {{ "{:.0%}".format(s.cumulative_pct) }} |
Average cost per file change by session length.
Which models Claude Code uses and their relative cost.
When you start focused coding blocks. (?)
Day-of-week work patterns.
Extended thinking usage — sessions where Claude reasons before acting.
| Session | Thinking | Errors |
|---|---|---|
| {{ s.custom_title or s.started_at[:10] }} {{ s.project_name }} | {{ "{:,.0f}K".format(s.thinking_chars / 1000) if s.thinking_chars >= 1000 else "{:,}".format(s.thinking_chars) }} | {{ s.tool_error_count }} |
Which permission mode you use across sessions.
Most common tool-to-tool transitions — the workflow fingerprint.
| From | To | Count | Pattern | |
|---|---|---|---|---|
| {{ seq.from_tool }} | → | {{ seq.to_tool }} | {{ "{:,}".format(seq.count) }} |
{% set max_count = tool_sequences[0].count %}
|
How long you take between Claude's response and your next message (excluding auto tool results).