Cauza: am rulat `find transcripts -size -1k -delete` cat transcribe.py scria
in acelasi director. faster_whisper scrie bufferat, deci o transcriere de o
ora sta sub 1KB minute in sir; find a sters fisierul de sub proces, Python a
logat "Transcribed" pe un inode disparut, si audio-ul a fost sters. Lectia
11 (25.11.2025) pierduta complet, re-descarcata.
- transcribe_file() intoarce bool, verifica nr. cuvinte vs durata audio
- pull_moltbot.sh: staging + mutare doar peste 1KB, in loc de find -delete
- PROGRESS.md: documentat, ca sa nu se repete
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MtSTyTmt6AbCL9j5ajEDm1
