Volume 1 of the Edge AI Deployment Pitfalls series. The official path works — until it doesn't. Your AIMET model is rejected by a compiler flag you never set. The default checkpoint quietly splits the wrong precision. A 9 GB vendor asset freezes at 70% and the downloader calls it 'slow'. A bias tensor nobody quantized waits until link time to explode. This volume is a field guide to the Qualcomm QNN / AIMET failures that actually happen when you deploy GenAI models to QCS6490 hardware — every entry drawn from a real run, with the exact error string, the root cause, the fix or workaround, and the evidence (Hub job IDs, upstream issue numbers, file hashes, on-device measurements). Later volumes cover RKNN conversion pitfalls on RK3588, QCS6490 deployment health checks, and measuring edge AI honestly. Written by Edge Model Lab, from the team behind the public rk-bench and qcs-bench on-device benchmark leaderboards.