Hisashi Space

During August I found that emails with my own email domain got into the spam folder in Gmail no matter how I made a delivery. SPF, DKIM and DMARC verification passed smoothly and the domain was not on any blocklist.

I tested four kinds of sending a message with 'test' only and no links: first from my Workspace, then SocketLabs, an old hosting email server, and finally via Gmail API. All four resulted in spam. Consequently, I believed the issue was related to Gmail being against my domain rather than it being the email route that caused the problem.

The situation was dramatically changed when people started replying. I launched a birthday gift sign-up where readers were required to reply to the confirmation mail, and they literally fetched the mail out from their spam folder and replied. Every day I sent a test message to myself, and since September 1st it has landed without problems in the inbox every day.

I run 8 or more sessions at once, so terminal hangs and crashes show up way more often than I'd like. claude --resume looks like it should help, but it doesn't always pull the whole history back. I've stopped relying on it.

Our workaround is small. Every time the agent finishes a reply, a Stop hook writes a little card for that terminal pane, holding my last instruction, the tail of the agent's answer, the working folder, and the path to the full session log. It's a plain text file that gets overwritten on each run, so it costs zero tokens, and cards older than 7 days get cleaned up on their own.

Opening a new session in the same pane or folder triggers a SessionStart hook that reads the latest card and drops it into the agent's context. That way the new session already knows where the last one stopped, and it can always go read the full log when it needs more detail. Usually that's plenty.

I believe the key way where AI assistants really become a big help from now on depends on how many tasks you can delegate to them in parallel.

When I get down to development work, I usually have eight windows open, and occasionally when the administrative tasks start pouring in I have thirty-plus running.

You see, AI is very swift and with advances in every AI model, it's possible to entrust an AI with longer tasks. Consequently, if I only had one session open, I'd be the one waiting on the agent. Several at once is different. The agents are the ones waiting for our decision while the work continues and I believe that is how we get far more results done.