Spend report · @zen-vibecoder-7
$710/mo burn
ranked against 38 devs · median $473/mo
1.5×
Burns 1.5× the median dev — $710/mo vs $473. Top 37% of the board.
96%
96% of tokens re-read context already loaded; only 2.37% was newly written code. Paying to reload, not to create.
68%
Opus drives 68% of the bill — the single biggest cost lever on this profile.
▲
Last 3 days ran 2.04× the personal norm — the signature of a runaway loop or caching bug.
Where this landspercentile vs cohort · ◆ median
Monthly spend$71063rd pctl
Tokens that were actual code2.37%63rd pctl
Spent re-reading context96%95th pctl
Opus reliance68%58th pctl
Burn per active day$2461st pctl
Model mixby cost
Opus 68.3%Everything else 31.7%
Cut the burncomputed from these numbers
Route routine work to Sonnet — saves ~$116/mo
Keep Opus for architecture and hard debugging; /model sonnet for edits, tests, boilerplate.
Cut context re-reads — saves ~$85/mo
/compact at natural breakpoints, /clear between tasks, trim CLAUDE.md — every turn re-reads it.
Audit the last 3 days for a runaway loop
A retrying agent quietly eats 10–100× a normal session.
Cohort distributionprojected / month
<$100
5
$100–300
11
$300–600
7
$600–1k
you
$1k–2k
5
$2k–5k
3
$5k+
2
$ npx tokenrot
Out-burns 63% of developers. See where you land.