The Future of Visual Studio: Interactive Expert Panel | What’s Next for Visual Studio
Watch on YouTubeVideo summary
The panel discussion outlines a strategic shift for Visual Studio centered around the integration of the Copilot SDK as a primary method for enhancing consistency and bridging gaps in developer workflows. A major focus is placed on enabling seamless movement between different application surfaces when utilizing AI features, while simultaneously investing heavily in fundamental improvements such as performance optimization and stability. These foundational upgrades are crucial for supporting advanced scenarios like opening massive repositories containing millions of files without causing system slowdowns, effectively allowing users to manage large-scale projects or multi-repo environments more efficiently through enhanced "Open Folder" capabilities that reduce excessive indexing times.
Beyond AI integration, the roadmap includes a continued modernization of the user interface and experience, with plans to gradually convert remaining areas to Fluent design principles for heavily used sections. The team also addresses the evolving role of Visual Studio as a base for tools like SQL Server Management Studio (SSMS), acknowledging that building upon the main IDE provides a richer experience than previous standalone approaches. Furthermore, the discussion highlights a critical need to improve code review processes in an era where AI agents can generate massive change sets instantly; new tools are being developed to help developers confidently review, validate, and take ownership of complex code generated by external agents or local CLI interactions.
To support these advancements, Microsoft is introducing more flexible options for AI access, including the ability for users to bring their own keys or utilize locally deployed instances like Azure Foundry, which is particularly beneficial for organizations with specific security requirements such as federal government entities. The rollout of the new Copilot harness will be handled carefully, starting with internal early previews before a broader release, and the team is actively investigating ways to allow users to switch between different agent modes without restarting the entire application. This incremental approach ensures that feedback from the community directly shapes the product evolution, maintaining a balance between embracing new AI workflows and continuously refining the core developer experience.
Ultimately, the vision for Visual Studio's future involves adapting to changing developer workflows rather than simply adding more features in isolation, ensuring the platform remains relevant for the next quarter-century of development. The team emphasizes that while the focus is currently on Copilot and AI-driven capabilities, this does not preclude improvements elsewhere; instead, it reflects an evolution where the core investment supports new ways of working. By consistently incorporating suggestions from the developer community and releasing updates based on real-world usage, Visual Studio aims to stay ahead of industry trends, offering a robust environment that can handle both traditional coding tasks and the emerging complexities of AI-assisted software development.
Read the full video transcript
What are you guys focusing on now and
kind of the Q3 Q4 roadmap that you
expect to to ship over the next year?
>> Yeah, I mean
I mean I mentioned one big one, right?
Which is we will move to the Copilot SDK
as being, you know, one of the primary
ways that we try to make you know, close
that gap and make things more
consistent. The other thing it does is
something else I I mentioned a little
earlier, which is
you know, we're we're focused on
essentially making it possible to move
between these surfaces very easily if
you're using Copilot, right? So, that is
also a pretty big focus. Now, I'm more
at the fundamentals level, just Visual
Studio in general, right?
We continue to work on things like
performance and, you know, stability.
Um
in order to enable some of the things
that I mentioned, right? Like
you will see you should see
uh things like open folder become a lot
more capable.
Um and
that can mean that for very For example,
for very large, you know, repos, right?
Or even multi-repos, you know, being
able to open those
in in if if you're having trouble like
you don't have a solution that covers it
all or you want to like open something
broader, you'll have that option now in
in in open folder.
Um and part of that has to do with like
you know, more of the the like we open
folder right now, it does substantial
amount of indexing to make things later
faster.
Um
but you know, if you open something that
has
you know
10 million files in it, right? Then that
indexing will peg your, you know, CPU
for a good amount of time. So, we're
allowing for more configuration in that
area.
Um we continue to invest in sort of
trying to make the general UX and you
know UI
a lot
more modern
looking. The problem I always have when
I say modern is that everybody's like
definition of modern sort of changes on
you know a pretty
regular cadence.
So we will keep up with that
and you know if you don't feel like
something looks particularly modern
please
please give feedback.
Um
we haven't fully converted every single
piece of UX and UI within Visual Studio
to
you know the fluent
sort of design. So we will continue to
do that for places that are heavily
used.
Um
and then
uh
we have
I mean we talked a little bit about the
idea of Visual Studio being used base
for SSMS right? So there'll be some
pretty substantial
um improvements in SSMS in general.
Um it sort of went one way for a while
and then decided that actually building
on top of Visual Studio was going to
provide a richer more capable experience
and so
um they keep doing that and so we will
do more work
uh in that area.
Um
I mean there's some other things that I
don't know if anyone to add or think so.
>> Yeah I mean I think what we're focusing
on is we're helping developers review
code like the idea of like hey obviously
code generation has got drastically if
not exponentially cheaper now but the
the actually the the risk of it not
working is not necessarily changed if
not that's become even harder. So you
know we appreciate that like as you use
AI to have it directly generate entire
change sets right? You have an agent
that goes away comes back with huge
change sets or even if you're just
working locally with a CLI you know
you've got one of the worst problems
right? We have to review someone else's
code. And so we want to remake the
experiences for that much better Visual
Studio so you know you can more easily
review code and in a sense to make sure
it works for whatever system you've got,
meet your requirements, stuff like that.
So you can, you know, be much more
confident about checking in that code
and taking ownership of it. So both in
the outer loop case where there's other
people's code, like your traditional PR,
but also just like, "Hey, I had an
agent. It went away, burned several
million tokens, and it's got 500 lines
of code. What the hell is this?"
So it tools that help with that,
including fundamentals, which I think
you've just seen in Git, like we added
Git sub module support
and work trees, stuff like that. So, you
know, they all build to make a better
experience.
>> So you mentioned your improvements that
you're working on with Copilot in Visual
Studio.
>> Yep.
>> And I know that's probably a massive
task.
>> Uh it's not small, yeah.
>> Yeah. So is that something you plan on
doing like incremental improvements?
You're going to chip away and get this
feature updated to the SDK and then chip
away and do that? Or are you going to do
more of a big bang?
>> Uh it's it's almost
uh it's almost the third a third option,
which is that we we did the a smaller
version of it through agent the agent
preview choice, right? So if you do
agent preview, it's implemented as what
we would call a responder, but it's
basically a an agent itself, right? So
you select
uh it explicitly, and that is running on
the SDK, and that will provide you an
experience, you know, that should be,
you know, much more uh like similar to
what you'll see in the future when we
move the entirety of it. Now, the
entirety of it is what we call a
harness, right? Which is like
fundamentally underpins the entirety of
all the other agents, all the other
features in Visual Studio that are
leveraging Copilot. That is more the big
bang, right?
Um
big bang, I don't know if I love the
term cuz it makes it sounds like it is
exploding or something, but like
uh you know, you know what? It it it
should be um
it should be better. Now, we will roll
that out more carefully, right? Um
but
uh we we actually have
the the we have the early versions of it
available internally. So, if you
are using internal, you know, like the
the
early early preview bits, right? Then,
you'll see that much sooner
than than others. And Taysser wants to
add on to that.
>> We have always been building with you
guys, and so your input is very crucial
to us, right? And so, whether it is the
agent preview that folks were mentioning
or even the new harness, you'll be able
to test that as we are rolling it out,
and we'll definitely keep an eye on the
feedback, and you guys will be able to
steer as we are improving things. So.
>> Yeah, one thing so the agent the the
nice thing about the agent preview right
now is that, you know, you can switch
between. So, if you if you, you know,
agent preview isn't giving you good
result, you can switch to agent and vice
versa, right? With the harness, the way
it currently works is you have to
restart Visual Studio. That's an area we
might
see if we can investigate to try to make
it better so that you could
switch more easily between. Um
but at the moment, it'll feel more like
it's holistic as opposed to
smaller.
>> So, would it be fair to say, besides the
obvious Copilot innovations and work
being done there, as Visual Studio as a
whole, would it be fair to say that
you're really focusing on stabilizing,
improving performance, and basically
Visual Studio is kind of
Do you believe it's reached like a
critical mass of features to where
you're not seeing necessarily the need
to add more features to it, and possibly
even even taking the path of SQL Server
Management Studio, where maybe these
big,
you know, parts of the application
Visual Studio that aren't used a lot
maybe pulled out into a standalone
reusing the same shell?
>> Yeah.
So, I'm sure Andy has So, this is going
to be a like an opinionated answer,
right? Like in the sense that I have my
opinion and then other people will have
their opinions, too, right?
I I wouldn't
No, I mean I don't think that we're done
uh with Visual Studio in terms of the
capabilities or features that are we're
adding, right? Like we we're definitely
focused more on the Copilot side
uh at the moment. Um but in order to
enable that, obviously, we have to
invest in a bunch of other areas, right?
Which are going to add
capabilities that are good, right?
Regardless of whether or not you happen
to be using Copilot, right? Uh or any
AI. I mean, I guess we didn't mention
this, but I I should mention one other
thing cuz it may be interesting to a
bunch of people here is that we're also
making it so that you can use bring your
own key, right? Like so, basically, you
don't you don't need to necessarily
buy into GitHub Copilot if you don't
want to, right? Like if you you for
example, you work for the federal
government, and that's not an option,
right? But you could have a locally
deployed to like, you know, Azure
Foundry or something like that, right?
Like an Azure local instance. Then that
is will become an option.
Um but no, I wouldn't I wouldn't say
that we're We are looking for
opportunities for improving the
developer Now, the the thing is the
developer workflow and experience,
right? Is is changing.
Um and so, we want to optimize for what
that looks like with Visual Studio in
the future. I mean, Visual Studio, like
I said, I've been working on Visual
Studio
my basically my whole career, like
25-plus years, right?
I hope that the folks in this room that
have you know, that work with me, that
haven't been here 25-plus years are
working on it 25-plus years from now.
Uh and for that to be true, we can't not
invest in new workflows, new
capabilities, new features, right? And
so
that is what I want to do.
Um
the set that we do invest in, right?
That where is about I mean it's
evolving, right? Like so there's some
that I mentioned, right? That I think
will be, you know,
hopefully
really useful. But the other thing that
is again super useful is as you, you
know, are doing development, you're
like, "Ugh, I really wish that X,
right?"
Let us know.
Uh that is an area that we will look at.
>> Yeah, I mean just to double down on that
last point, we continue to watch the new
developer community where people like us
at Epic Games House can upvote and make
suggestions. And if you look at the
releases, we consistently do release new
developer community suggestions. So
that's going to continue to be a thing.
But obviously the the majority of work
is like, "Hey, the developer workflow is
changing and we need to adapt to that."
But that's not to say like we're not
going to continue to make improvements
elsewhere at all. Just that our focus is
really on, "Hey, embracing the moment."