The actual next big thing? Not a language. Not some hardware change ( I mean, ARM servers could be big, but only the server people will care. It won't be the Next Big Thing.) Not block chain - that's close to played out. Not machine learning - that's too far along. It's what's now, not what's next.
I see two possibilities. It might be quantum computing. I doubt it, but it might be. It also might be wearable VR. Again, I kind of doubt it, but I doubt it less than I do quantum.
The reason new frameworks can't dislodge the big 3 is because they aren't bringing new ideas. Any framework that changed UI development as much as React did would absolutely be worth noting.
I think the next big thing will be Capability Based Security -- the good kind, not the "allow location access" kind that people confuse it with.
I also look forward to the resurgence of the personal computer.
All known "quantum error correction" algorithms are focused on not flipping qubits, and ignore phase.
For instance, using Grover's algorithm to crack a 128 bit key requires 2^64 quantum operations in sequence. To make this happen your phase shift and interaction between the qubits has to be low enough that it doesn't accumulate to the level sufficient to flip a single qubit in all that time.
You'd need signal to noise ratios of about 380 dB to make this happen. This would be 220 dB more!!! than detecting the signal from the Voyager (10^-20 watts) probe right next to a SuperNova explosion (10^40 watts)
Stable Diffusion is already making waves in digital art. I expect more specialized models fairly soon.
Free data and services we used to take for granted like YouTube, farming software, and car seat heaters are becoming forcefully monetized. Will we someday refer to these as the good old days when search engines were “free”?
Solar
I think as people get more knowledgeable of tech they will care more about security or privacy. It is already a big selling point of messaging applications.
There are some signs Rust is heading in this direction.
Efforts like Dafny are also pretty interesting, but still far from mainstream.
I think a large codebase (20-40 KLOC) is about the maximum one can routinely verify in Dafny before stuff gets too hard. And coding speed is pretty slow.
Another benefit of CUE over Dhall is that I can use CUE from Go. Very few people use Haskell, Go is an industry standard at this point
A way to escape the google silos and protect your private data from mass surveillance, but without losing the usability. Things like locutus, solid, etc.