AcqualyticsAcqualytics
Engineering · AI leverage

AI usage & agent loops

Every AI job is a unit of work made of loops (agent iterations). We tie tokens and dollars to shipped value — not just raw usage.

avg · 90 days
Tokens / shipped feature
2.50M
avg · 90 days
Loops to merge (avg)
4.3
agent iterations per merged job
AI-authored
58%
of merged changes
Spend / week
$600.67
8.9 features/wk avg
More effective AI

Before Acqualytics → now

after assessment + remediation
Tokens / feature
4.6M→2.1M
less waste per shipped feature
Loops to merge
7.6→3.6
fewer agent iterations
Features / week
0.7→10.6
more shipped value
Value, not just usage

Tokens & cost against features shipped

per week
Tokens (M)FeaturesCost ($)
The loops

Recent AI jobs

job → loops → outcome

Form A grid migration

merged
claude-opus-4 · nwg-quotefix#812 · 3 loops
Tokens
1.80M
Cost
$41.20
LoopTokens inTokens outFilesTests addedResult
#1620.0k160.0k146partial
#2480.0k130.0k98partial
#3320.0k90.0k45pass

Unified SSO login

merged
claude-opus-4 · nwg#1204 · 2 loops
Tokens
1.24M
Cost
$28.40
LoopTokens inTokens outFilesTests addedResult
#1540.0k150.0k114partial
#2440.0k110.0k69pass

E2E test pipeline bootstrap

merged
claude-sonnet-4 · nwg-quotefix#787 · 3 loops
Tokens
930.0k
Cost
$12.90
LoopTokens inTokens outFilesTests addedResult
#1380.0k120.0k2240partial
#2220.0k60.0k822partial
#3120.0k30.0k310pass

Advanced search StarRocks load

in_progress
claude-opus-4 · nwg#1230 · 2 loops
Tokens
1.40M
Cost
$33.00
LoopTokens inTokens outFilesTests addedResult
#1600.0k170.0k123partial
#2500.0k130.0k76partial