[tunesmith]

Here are a couple of stabs at boring answers for commodity developers, at the risk of being wildly wrong:

For the next 5 years, any team that is big enough to have a collection of multiple restful endpoints and services, GraphQL is just generally superior. Federated GraphQL is worth investigating as well. For performance, look into the various protocols that are superior to json over http.

For frontend, if we're looking at a 5-10 year timeframe, I think it's still the wild west. I still suspect that React, while a good evolution over what came before, is still something of a conceptual impedance mismatch. Whatever model is best for reconciling UI concerns with network communication data flow, I'm not sure we've really identified the right primitives to design on top of. We're still rewriting to find the lower-level approach that really fits.

For AI, we're really distracted. We're on this road to engineer more and more convincing simulations of intelligence without really tackling intelligence, and we're going to eventually realize it's asymptotic and will hit another dry spell. When we get back to focusing on not how to make computers smarter, but instead how to use them to make people and societies smarter, then we'll make fast progress again. But this will require some political realignments. I think this is more of a 20-25 year timeframe. You can make a lot of money in the next ten years if you get into AI with the intention of finding even more effective ways to trick or con people in the short term, but the overall impact will be damaging.

Finally, I have no experience in this arena, but I think VR/AR/MR is going to be absolutely huge. Developers are going to have much more opportunity to play with this stuff than with machine learning, which will be part of it but at the lower level of libraries and device firmware. Integrating that tech with development tools, end-user experience, etc - that will be massive.

[jschrf]

I think GraphQL goes away in 5-10 years and we can place it next to the whole NoSQL craze and just move on, but that's me.

I've never seen it be anything but accidental complexity, but would love any concrete examples of where it's been worth the effort.

[tunesmith]

From the client perspective, I've always seen the main three benefits (compared to rest) as: avoiding overfetching, avoiding underfetching, and type safety. Ideal case is one query to get exactly what you want, and to know what you're getting.

[satvikpendem]

I love GraphQL, it makes writing typed APIs for multiple clients very easy, since client code can basically be auto generated from the GraphQL schema. More concretely, it forces you to think about your data, unlike NoSQL where you can just chuck a JSON blob of whatever you want.

[blovescoffee]

What are some of the protocols superior to json over http?

[rickette]

gRPC