Anthropic's dynamic workflows gave me the confidence to actually extend agent autonomy
I am way more willing to let agents run loose when the workflow is repeatable and recorded
built a suite last night: autonomous + human-gated skills across ops, search, build, and design
the autonomous ones gate on my Obsidian kanban boards
Claude and Codex doing the heavy lifting as a team
@8017@thsottiaux came to post this, then saw the reply and found it finally.
maybe make this more obvioous @thsottiaux ? Chris & I might not be alone
the single best way to stay focused while the AI frontier moves at a million miles a minute:
Keep your tinkering locked to your problems.
If you can't immediately see why it'd be useful to you, let it scroll on by.
context switching always has a price
@omarsar0 It’s becoming my default way of ingesting research, workflow logs, and strategic counterpoints
Fun how well getting a few good formats you like stacks well
you wouldn’t know it by looking at the timeline, BUT -
Claude Code and Codex play extremely well together if you just ask them to - “play to your strengths”
This is such a cool comms focused setup, thanks for writing this up!
I’m thinking about how to handle more childcare tasks and your type of flow form-fits pretty well to that (so many messaging apps).
Btw, I’ve been tinkering with Cora as an agent in my CLI - Super helpful for all things email mgmt, could be worth a try