Where do you talk to Amp most: terminal, editor, web, phone, Slack?
Mostly the iOS app on my iPad. I actually haven’t used my MBP for Amp stuff in the past week. Pretty much all iPad.
Favorite mode, and when do you switch?
Mostly medium or high. Ultra if I have an important an well defined prompt. I’ll explore Raw models from time to time mostly to see how open weight models compare to GPT and Claude.
Do you still develop locally? If so, when?
Nope.
What’s a workflow with Orbs that you absolutely love?
Having the agent capture a video of its changes.
How do you review changes?
Usually try to have the agent explain which files contain which hunks of changes. Then manually review the ones I determine are the highest risk. I usually try to get the agent to slim down its changes if I notice a big difference somewhere I don’t expect. Generally just try to keep commits at the “I can comprehend the number of files changed” level.
Which skills, plugins, or MCPs do you use? Anything you built for yourself?
For my Amp setup it’s pretty vanilla. Whatever is in the amp repo. On my westpac project I have a knowledge graph setup and some skills for orchestrating workers. These skills are highly tuned for code modernization. I think it makes more sense to have a specced out setup when you have a specific, verifiable task like code migration.
How do you use Puck?
I’ve used it for bug triage and interaction with slack but not very heavily. I haven’t really adopted it fully. If I want a meta-agent I will usually just have a normal thread that spawns off others.
What’s a prompt or phrase you keep reusing?
What do you still do by hand, and why?
Writing stuff that people will read.