An AI subscription detector. Its own promise: AI powered subscription detection from your bank statement. Discover hidden charges and save money. You upload a PDF statement, a language model identifies the recurring charges, and you get back a report on its own private link. One payment, no subscription.

Designed and built in code, with AI as the build partner. No design file, no specification, no handover between design and development. The AI pipeline and the interface were built together, not by separate teams in sequence.

• Design and development were the same step.
• When the model misreads a merchant, the user corrects it and the correction feeds back into the system, with no ticket and no release.
• One person covered positioning, AI pipeline, design and build.

Product, payment flow and correction loop in production.

17 days of effort across 41 active days. 367 commits.

• A detection model that combines a language model with hard coded rules. The rules handle what can be resolved without AI, so the model only processes what is left. Less time per statement, fewer tokens per run.
• The full site, responsive, with one time payment.
• The PDF reading and detection pipeline, with a private report link per user.
• User correction of misread merchants, feeding back into the system.
• Comparison pages against Mint, Monarch Money, PocketGuard, Hiatus and WeMoney, built for search and growth.