I open new session and start working on projects that process images of stuff I pilled up over the years and create ebay listings. Once I have about 5 to 9 listings I review them and have agent poste them on ebay via API. Trying to go minimalist lol. I top of at about 7 open sessions at one time. OpenCode has decent orchestration opportunities so I'm playing with that to see if what else I can automate so I have less mind numbing stuff to deal with and just to fun thinking. And when I fully maxed out my limits on all subscriptions and APIs are chugging on low cost "re-tries" I go for walk or bike ride, enjoy some sun and birds signing (or armin van buuren in my BC headphones)

sounds like a plan haha. How is OpenCode vs Claude Code orchestration? have you ever needed to modify the orchestration since it's open source? do you mind sharing more about your experience?

Orchestration is a lot of skill calling, templates (usually YAML) and gate checks. And the workflows, oooh lord the workflows. From my experience, if you get the agentic instructions right, the workflow runs fine on OpenCode, Claude Code, Codex and Antigravity. I have to tinker around since for personal stuff I use only subscriptions. For business I run the API, but that gets expensive very quickly, especially with lots of agent runs and validation cycles. That said, you test your agentic instructions in OpenCode and then port the agent logic into a framework like LangGraph or CrewAI and run everything on Amazon Bedrock AgentCore. It can take an afternoon to get this set up for production, but it's nice to have something running on its own with minimal intervention. You can do controls and output capture over MCP, the API, or even S3.

Mostly I try to understand the code around the change. If I cannot review the diff afterwards, the agent did not really save me much time.

you wait for it to finish then review or give feedback in the meanwhile? if it's the former, i'm curious about what you still do in the meanwhile