Beau Simensen
@simensen
Staff-plus Software Engineer
- Madison, WI
- beausimensen.com
- github.com/simensen
- Sponsor on GitHub
- Joined October 2026
Commit activity
16 commits in the last year
longest streak 4 days
441 / 1,000 XP559 XP to wheel
Tue, Nov 3
Mon, Nov 2
replying to @simensen
Alright, seems like that might be the trick. Just strange that the at itself says it isn't there but after a few seconds post-post it resolves. @scinddev @kybernata
replying to @simensenShow thread
Going to try to mention them to see if that helps? I'll start with @procrastivity first and then maybe try some of the others if it works.
Sun, Nov 1
replying to @paulinevos
Yeah, it's a lot more work and I find either extra coordination takes a lot of time or less time but things get missed. Especially if I want to use a specific mode for implementation and another mode for review. So much extra tune coordinating what changed, packaging the changes,…
replying to @mnapoli
Yeah. I haven't figured out if this is due to codex vs opus 5.5 or the fact that everything appears to be funneled through amp servers adding an extra hop?
replying to @michaeldyrynda
Sadly, where I think I could best leverage orbs is at work but we're restricted from working with external systems, so I can't really prove for myself if I'm right about that. :-/
replying to @michaeldyrynda
I suspect the people moving to orbs and loving it either came from a heavy git worktree workflow or would have benefited from a heavy git worktree workflow but never made the jump.
replying to @michaeldyrynda
It's a genius move on several fronts. I'm trying to incorporate a lot of the ideas into my own tooling. But "every thread starts from a blank slate" seems to lend itself more to product/feature work for narrowly scoped impact. Doing research and planning that sprawls far and…
Sat, Oct 31
replying to @paulinevos
I tend to like doing a lot of child threads. Coordinating that between explicitly separate hosts (orbs) is far from trivial. And because you can't charge agent mode for simple subagents, child threads seem to be the only option for delegating to a different agent mode/model. And…
I’m guessing you’d just have to specify you want to to base it off the current state irk and then keep telling it to rebase
Doesn't the agent itself orchestrate droppoffs of changes pretty good?
Or is that slow for you/not working well?
I feel this as well, but haven’t used orbs yet. I do wonder if this is less “it doesn’t do what I want/how I want” and more “I should try their way”
The friction is mostly from it being different to how I’ve always done things, so gotta be open to the possibility orbs might…
That's @simensen's first reply.