Загрузка видео...
Не удалось загрузить видео
Local machines had become a bottleneck for agentic mobile development at Shopify, so we gave agents their own remote dev environments. They now author 61% of our merged PRs 🧵
38,979 просмотров • 8 дней назад •via X (Twitter)
Комментарии: 29

Also, anyone can now do mobile development from Slack and use the app live in a browser. No Mac needed. Nothing to install.

Blog post

@Shopify Dream mobile dev env 😍

@Shopify Local dev and work trees ain't it Cloud was always the destination

@Shopify Very cool. I love the updates from Shopify

@Shopify When a few agents are working at once, how do you keep each browser preview tied to the right branch?

@Shopify Each agent session is isolated from the rest so it gets its own preview

@Shopify Great stuff! The future of mobile development and cutting edge technology is being developed at Shopify 🚀

@Shopify Wow 🔥

@Shopify the live app in the browser is the part i'd lead with for the rest of the team. 'try this screen and tell me what's wrong' is a much easier handoff than 'check out this branch and get a simulator running'.

@Shopify 61% of merged PRs is wild. When I built my browser agent with Playwright I learned the sandbox matters more than the model — an agent without its own environment just makes confident messes on your machine.

@Shopify Giving agents their own environments seems like a useful step beyond individual coding assistants. Which made the bigger difference in practice: isolated builds or access to devices and simulators?

@Shopify this is so good! how fast is it?

@Shopify Startup takes a few minutes but we are optimizing it aggressively. Our goal is to make it instant.

@Shopify awesome work! i’m looking forward to reading some blog post on how it all works

@Shopify

@Shopify This is great. One small thing missing is a way to also test on a physical device.

@Shopify We have that too

@Shopify I love it, that's very cool! What about testing features that require retail hardware integrations in the POS? Is that still a bottleneck for AI automation?

@Shopify Haven't tackled that yet, but soon

That's so good and it’s what we’d LOVE to do too. As the pudding now has the proof, Shopify should build the infrastructure to enable us to be able to launch extensions and apps outside of Shopify's platform so we can do the same thing... We have the app mock bridge but it's still not 100% like within Shopify. We want what you have 🙏

@Shopify How's your review process for these prs?

@Shopify Agents review all prs first and iterate until everything they find is resolved. Humans do a final check before merging.

@Shopify Nice! I've found AI keeps finding new bugs every time time I ask it to find something. Has that happened to you too?

@Shopify Yes, there's definitely a point of diminishing returns if you let them keep going. We have checks to stop them from nitpicking endlessly.

@Shopify Which architecture are you using, MVVM or MVI? I imagine they modify the state and interact with actions to achieve development speed

Keeping each agent's preview separate makes a lot of sense for human review. You can try one change while the other agents keep working

@Shopify woah, great stuff! any issues managing env vars or keys remotely? also, i'm just imagining a rack of mac studios in your home office 😂

@Shopify It wasn't too bad. We've abstracted all that complexity away though so the team doesn't have to worry about it. Rack of mac studios in my home office is the dream 😄

