increased interactivity: Haskell is moving towards a more interactive workflow. We recently got deferred type errors which let you turn type errors into warnings. This means that you can load and interact with code that doesn't typecheck, making the repl much more useful.
The next step after this is the addition of type holes. Type holes let you leave a hole in your program and the compiler tells you what type it expects for the hole. This is very useful if you get confused or are just working on something complicated. The feature was borrowed from Agda and the general idea is to let you develop in a "goal-oriented" way: you can develop step-by-step where, at each step, you know what types you already have and what types you need. Your task now is just to fill in that gap.
In essence, the goal is to turn a more batch-oriented experience into a conversation with the compiler. You can expect some editors to start taking advantage of these features, behaving more like Agda mode.
improved tooling: Recently, there has been a greater push to improve the general Haskell ecosystem. A new version of Hackage has been rolled out. FP complete was founded and released a commercial web-based IDE. Cabal has been noticeably improved including native support for sandboxes.
I expect this trend to continue, including improvements to popular Haskell editors like Emacs. I mentioned increased interactivity earlier, which is at least partly an editor feature. I also imagine that somebody is likely to add Haskell support to Light Table at some point, although that might take a while yet.
One particularly exciting project to point out is IHaskell. This builds an improved Haskell repl on top of IPython, enabling richer output like images or HTML. It's still somewhat experimental, but getting there. It plays well with visual Haskell libraries like diagrams and Chart.
more uptake: functional programming has been getting more and more press lately, and Haskell has definitely benefitted. I've certainly met an increasing number of people in the valley who are interested in using Haskell for projects or startups. We've certainly seen a marked increase in the number of people coming to meetups and hackathons, which is a good indicator.
I don't think Haskell will become super popular in the near future, but I do think it will see more use and be less marginalized.
I really can't make many predictions about the long term. Haskell is still going strong and I don't see it losing steam for a long time yet. There is still plenty of research to do and people are still interested and willing to work on the language itself and its ecosystem.
The Haskell community is not the largest, but it's extremely active and productive. Haskell is still going to be well supported and actively developed for a long time, so it's a good choice for a long-term language.