Weekly report

Week of 2026-04-27

This report is private to you. On auth-svc: shared with the project lead (set at project-join). On billing: not shared (your default). Change anytime in /settings.

14
Sessions
87
Prompts
2
Writeups published
Where time went

Worked on

Project-by-project breakdown of where this week's sessions landed.

auth-svc6 sessions

credential rotation work

auth-svc4 sessions

test fixture cleanup

billing2 sessions

invoice retry logic

general2 sessions

smaller one-offs

Spend

Token usage & cost

A cross-session rollup of what your Claude work cost this week, broken down by model with cache-hit and output:input efficiency.

over budget≈ $42.18 this week — above your $30 alert threshold. Set your own ceiling in Settings → Cost alerts; sessions get a nudge at 80% and a red flag at the line.
This week · all sessions
Could improve$42.18 est.
/compact saved$11.94 avoided across 6 sessions by auto-compact in this window.
$42.18Est. cost
71%Cache hit
0.19Output : input
By model
claude-opus-4-8286,400 in · 132,900 out · 4,120,000 cache
$36.74
claude-sonnet-4-6198,700 in · 71,300 out · 980,500 cache
$4.62
claude-haiku-4-5412,000 in · 46,800 out · 690,000 cache
$0.820

Costs are estimated from per-model list prices. Cache reads are billed at a fraction of fresh input — a high cache-hit rate is most of why this number stays low.

Patterns

What's working & what to watch

Patterns Claude noticed across this week's sessions. Surfaced as observations, not as a grade.

✓ strong

Scoped, hypothesis-driven prompts on the credential work ("here's the error, here's what I think is happening, verify before changing code") — these one-shot consistently.

⚠ watch

Three sessions started with a paste of the full stack trace and no context. Claude's first reply was usually wrong; you corrected on turn 2. Worth front-loading the hypothesis the way you do on the credential work.

Coaching

Lessons that match what you did this week

Pulled from the team library. Read these — they're load-bearing for the patterns above.

active@alice

Front-load the hypothesis, not the trace

Pasting a stack trace alone gives Claude permission to guess. Stating your hypothesis first turns the turn into verification, which is cheaper to be wrong about.

active@alice

Scope rm operations explicitly — never trust 'old' as a filter

When asking Claude to clean up files, always scope by an objective signal (unreferenced in code, last-modified > N days, matched glob). 'Old' is interpreted aggressively.