A modern programming language that makes developers happier.

kotlinlang

<https://shopify.engineering/back-to-native|shopify.engineering/back-to-native>

But why Swift and not directly KMP? :thinking_face:

Mustafa answered that question in a comment <https://www.linkedin.com/feed/update/urn:li:activity:7503813416202018816/?dashCommentUrn=urn%3Ali%3Afsd_comment%3A%287503819662716014592%2Curn%3Ali%3Aactivity%3A7503813416202018816%29&amp;dashReplyUrn=urn%3Ali%3Afsd_comment%3A%287503849026929381377%2Curn%3Ali%3Aactivity%3A7503813416202018816%29|https://www.linkedin.com/feed/update/urn:li:activity:7503813416202018816/?dashCommentUrn=urn%3Ali%3Afsd_comment%3A%287503819662716014592%2Curn%3Ali%3Aactivity%3A7503813416202018816%29&amp;dashReplyUrn=urn%3Ali%3Afsd_comment%3A%287503849026929381377%2Curn%3Ali%3Aactivity%3A7503813416202018816%29>

I still think AI+KMP would be better

&gt; Shopify has been using LLMs to build software since 2021 (one year before ChatGPT!).
That's a strange flex. Like they are saying „we used AI first“. 

I doubt it. Before 2025 I found coding agents &amp; models not good enough for agentic coding at all.

Can't point my finger at it, but somehow Shopify makes a bad impression to me with that post, comments and their claims. Maybe they don't know what they are doing. :eyes:

Time will tell if that Helix system really prevents slop.

Maybe they still trust AI too much.

Going with KMP would be the safer approach in my world.

Even if AI maintains it you still have two codebases and you need a lot of extra tooling trying to control it. I feel with KMP we are past this and now they are going back.

This is not my area of expertise, but curious to understand what would be the benefit of a single KMP codebase over a single React Native codebase? Would it be worth the migration effort?

React Native means you have to deal with TypeScript/JavaScript and all its issues.
Using KMP you can use Kotlin, the best language there is.

They migrated to React Native at a point in time where Compose for iOS support wasn't there. Now it's here.

Here is more on that: <https://youtu.be/9WjPdgOK-68?t=4550&amp;si=J5YfnlyQ1aAMsuLx|youtu.be/9WjPdgOK-68?t=4550&amp;si=…>

I think it's important to remember that Shopify can afford to burn astronomical amounts of tokens, maybe not a fair comparison with "regular" companies. I think the kmp tradeoffs are more favorable when you have actual budgetary restraints

&gt; They migrated to React Native at a point in time where Compose for iOS support wasn't there. Now it's here.
Yes, if CMP existed at the time, they may choose it instead. The way things are now, the only people choosing RN already have a love for web tech, and people choosing Flutter don't know about CMP.

Of course, Flutter has a larger community and more libraries and widgets and 3rd party SDKs than CMP, but it started as a niche. The Kotlin community is larger than the dart community and can easily stand up to the challenge to make CMP &gt;&gt;&gt; Flutter

&gt; it's important to remember that Shopify can afford to burn astronomical amounts of tokens
100%! They also had enough $$ to invest in whatever Helix is (and probably other tools, experiments, context pipeline, etc) so that they could get the AI to do things right.

Touchlab was built an agentic tool to migrate native iOS -&gt; native Android. We had a client with an iOS dev using Claude to do the same. What took that dev 2 weeks to reach 70% feature parity, or tool took 2 days to reach 95% parity.

Touchlab spent a lot of time bringing down token usage because we don't have the resources of Shopify. I'd bet their solution is much less token-efficient than what we built.