Cold email

Two batches a day, eight hours apart. 25–30 is the range that works; the limit on the first is 50, and above that this stops being outreach and starts being bulk mail, which Gmail treats differently. The second batch of a day is capped at 20 and cannot start until eight hours after the first one finished — two full batches back to back read as one big send split in half, which is the thing the gap exists to avoid.

At once is how many messages are in flight together — 5 turns a 25-minute batch into about 4. Gap bounds how fast they actually leave, whatever that number is, so the batch is a steady stream and never a burst.

Finding people runs on its own clock, in the afternoon, so a failed search has the rest of the day to be noticed instead of eating the morning's send window. It drives the Chrome already signed in to LinkedIn inside the scraper container — watch it at localhost:7900. Each search spends a monthly allowance you cannot see, shared with the LinkedIn connect agent, so the company count is deliberately small.

Both clocks need the worker running on your machine: python coldemail-worker/run.py watch. Nothing sends, and nothing is scraped, while it is closed.