Gym App QR Check-In Rebuilt with AI
A developer used GLM 5.3 and DeepSeek v4.1 Flash to reverse-engineer a Flutter gym app’s offline QR check-in system, then rebuilt it as a lightweight SvelteKit web app that avoids the gym’s forced social feed. The process uncovered plaintext key material in device logs and validated the recreated AES/WebCrypto output against the original app. [Source](https://shayon.work/blog/reverse-engineering-gym-pass)
The real lesson is product design: unnecessary engagement mechanics can motivate technically capable users to build an escape hatch.
- –LLMs made reverse engineering accessible to someone attempting it for the first time, but human intuition still steered the investigation.
- –Plaintext logs exposed the most damaging weakness; client-side secrets remain vulnerable on rooted devices.
- –Round-trip crypto tests were misleading because the same implementation errors affected encryption and decryption.
- –Independent validation with WebCrypto, OpenSSL, known vectors, and real QR output prevented false confidence.
- –Forcing a social feed onto a utility workflow created friction without improving the core check-in experience.
DISCOVERED
1h ago
2026-11-05
PUBLISHED
4h ago
2026-11-05
RELEVANCE
AUTHOR
yewenjie