Debug Shop App

Penny winSelf-reported run (local Penny Compare)Oct 10, 2026

Testing more complex session

Claude Code 2.1.296claude-sonnet-5-5Native Penny

Cost−$0.242

−43%

Penny 43% cheaper

Native

$0.564Penny

$0.322

Tokens−516.6k

−58%

Penny 58% fewer tokens

Native

889.4kPenny

372.8k

Model calls−3

−19%

Penny 19% fewer calls

Native

16Penny

Session replayProvided with explicit permission of session creator

Claude Code

› In shop-bugs/ there is a small storefront (vanilla ES modules, no dependencies) and README.md holds the business rules as the specification. Using only Node built-ins (no installs): run node --test inside shop-bugs/, find the root cause of every failing test in src/ and fix it there so the code follows the README rules (do not weaken or delete tests), run node --test again until everything passes, then check the same rules hold in the UI logic in app.mjs. Keep shop-bugs/index.html working when opened from its folder with relative paths only, and finish with a short list of each bug and its fix.

NATIVEclaude-sonnet-5-5

read: 817.6k
write: 50.1k
out: 19.8k
cost: $0.564

PENNYclaude-sonnet-5-5

read: 318.9k
write: 37.1k
out: 16.5k
cost: $0.322

Native

shop-bugs/index.html

Loading page…

Penny

shop-bugs/index.html

Loading page…

Session Details

	Native	Penny
Model	claude-sonnet-5-5	claude-sonnet-5-5
Model calls	16	13
Cache read	817.6k	318.9k
Cache write	50.1k	37.1k
Fresh input	1.9k	283
Output	19.8k	16.5k
Total tokens	889.4k	372.8k
Background calls	claude-haiku-5-5 ×1, claude-sonnet-5-5 ×1	claude-haiku-5-5 ×1
Cost	$0.564	$0.322

Costs are API-equivalent estimates at Standard prices, including the main conversation, subagents and background calls such as titles and the auto-mode reviewer.