LOG Writing - field reports
Notes from production.
What held, what broke, and what was cheaper the second time - hands-on, from my own projects.
AUG 02, 2026AI INTEGRATION · 7 MINJUL 20, 2026AI INTEGRATION · 9 MINJUL 08, 2026AI ENGINEERING · 16 MINJUN 02, 2026AI ENGINEERING · 11 MIN
LLM-Personalized Email Campaigns with Python, Gmail, and Local AI
A local LLM turns CSV rows into individualized Gmail campaigns, with preview, throttled sending, thread-based reply tracking, and classification.
Implementation Lessons: Hidden Headaches of a Local Gmail AI Agent
The Gmail API details that quietly shaped a local email agent: history-based sync, label semantics, threading, and the failure modes each one hides.
How I nearly doubled my premium model capacity without paying more
Treating the most expensive model as an R&D department instead of a coder: it plans, a critic reviews, cheaper agents execute. With the measured ledger.
LLM as Router: Intent Classification for a Local Telegram Email Agent
A local LLM routes Telegram messages to intent handlers: slash commands first, natural language as the fallback, and only when nothing cheaper parses.
EXT Published elsewhere