I’ve spoken with a bunch of teams who felt they needed to switch to Python to build AI features, abandoning their normal stack to chase the ecosystem benefits.

In most cases, it’s not gone great. Learning AI while also picking up a new language and toolchain is a lot, especially when working with AI alone can be a big learning curve.

We’ve built all our AI at incident.io in Go and it’s worked brilliantly. Static typing + proper concurrency are exactly what you want when building AI systems, and you can get the ergonomics right with a bit of work (even in Go, a language notorious for being no-frills).

I’ve written up how we do it, with examples from our codebase. If you’re picking up AI and wondering if you need to learn Python too, know that for at least one team (us), sticking with what we know has worked out really well.

Link in 👇