Loading video player...
[Music]
Hello everyone. Thanks so much for
coming today. I'm Malavika and as you
heard I'm a product manager at Verscell.
I hope you've had a great day at the
conference. I've certainly been blown
away by what all of you have been
building with our open source tools and
infrastructure primitives. It's been so
nice to meet many of you in person for
the first time um and see some new faces
as well as familiar faces.
So today we've learned a lot about the
AI cloud. By now you're probably tired
of hearing about the AI cloud. You might
have be thinking, gosh, can I just get a
drink at happy hour already?
But just as a reminder in case the
message just hasn't quite sunk in yet,
Verscell is a unified platform for
building, deploying, and running
intelligent applications and the agents
behind them. Verscell's mission has
always been to abstract away the
complexity of managing infrastructure
and let you focus on building amazing
user experiences.
So when we think about how successful
we've been in abstracting away
complexity across the software
development life cycle, we've done a
really good job on making buildtime a
breeze. Framework defined infrastructure
means that you don't need to think about
the underlying infrastructure
primitives. There's no need for complex
orchestration or infrastructure as code.
We handle provisioning compute,
networking, caching and more so that you
can focus on your application logic. But
if we reflect on runtime, there's still
a lot of work we have to do to make that
as easy and breezy as build time.
Sadly, managing an application at
runtime is still quite a time consuming
effort for development teams. Like how
many in the audience here have dealt
with a major incident in the last week?
I I don't see that many hands up. I'm
surprised. I'm like, are you all using
GCP or something? Um maybe maybe you
there in the audience uh you're probably
feeling really good right now that
you're self-hosting Nex.js on a VPS in
Hetner. You're like, the cloud's not
going to bring me down. Um, but there's
so much manual toil associated with
incident response. First, teams need to
set up alerts and monitors to identify
potential problems. These monitoring
tools then in turn cause alert fatigue
as teams try to identify high signal
issues amongst all the noise. And then
once we identify an issue, teams spend
hours debugging the issue, trying to
identify the root cause and then apply a
fix.
So I actually here in the audience we're
going to make this a little bit
interactive. How much time do you think
developers spend on debugging an
incident? How many of you think it's
under 20% of incident time is spent on
debugging?
>> More 80 I heard 80%. I So we think it's
almost as who thinks it's more like 40%.
Uh okay, we've got we've and someone
thinks it's 60 70. Okay, that seems to
be maybe consensus. Um 50% actually
turns out that 50% of incident time is
spent on identifying the root cause and
figuring out who should solve it. Like
that's crazy. In 2025 with AI, we're
spending hours of valuable developer
time debugging issues. What if we could
reduce that to seconds?
Well, with Verscell agent, you can. Our
anomaly alerts monitor your application
for suspicious activities out of the
box. No configuration required. As soon
as we detect unusual behavior, Verscell
agent investigates the issue, performs a
root cause analysis, and diagnoses the
problem in a matter of seconds.
Unlike traditional observability tools
or infrastructure providers, we have
full context on your app. We built it,
we deployed it, and we're running it on
production. Even though we don't know
your app as well as you do, and
certainly not as well as your 10x
engineer, we're uniquely positioned to
give you an AI native approach to
ensuring reliability, performance, and
security at runtime.
Verscell agent investigations build on
top of our native observability tooling
which we've thoughtfully designed to
give you visibility into runtime
behavior with buildtime context.
Runtime logs give you granular
visibility into application behavior.
With runtime logs, you can trace through
the life of an HTTP request to your
application from the point in time that
it ingresses into the Versell network
all the way to the point that a response
is returned to the client.
We also provide you with opinionated
dashboards out of the box so you can
quickly understand the health of your
application, pinpoint issues, and
optimize performance.
With Verscell's recently launched
anomaly alerts, now you can actively
monitor your application for unusual
activity so that you can quickly
identify and resolve issues.
And lastly, with our query tool, you can
explore the vast amount of metrics we
collect and surface on your application.
You can author queries to answer a range
of questions like what bots are crawling
my application all the way to
quantifying the P90 time to first token
of various model providers you're using
in your application.
Versell agent investigations build on
top of all of these features, reducing
the need for manual exploration and
bubbling up key insights.
As we look to what's next for Versell
Agent, our goal is to ultimately
reimagine the way we interact with
observability tools.
Our vision is that the AI cloud will
repair and optimize your application,
not just inform you about issues.
Fundamentally,
we believe the AI cloud shouldn't just
present you with problems. It should
give you solutions, recommendations,
pull requests, and automated actions.
And that's the world we're building
towards with Versell agent.
You can get started with Versell Agent
today. As you heard, it's in public beta
and we're giving away $100 in free
credits to all Versel users to try out
these new features. We launched the code
review skill last month and that's
available to all Versel users. Agent
investigations are available all
starting today and they're available to
pro and enterprise customers who have
observability plus. And you can visit
the agent tab in the Verscell dashboard
to get started.
So I'm going to shift gears a little
bit. I actually want to spend time on
another important topic which is eval.
So how many of you in the audience are
using a dedicated AI observability
product for evals?
Also I see actually very few hands up.
Um interesting um but for those of you
who are you know AI applications are
non-deterministic.
So it's very important that we're also
monitoring the quality of the output.
agents daisychain a series of reasoning
steps introducing even more complexity.
And that's why we've seen a large
ecosystem of agent frameworks that build
upon open telemetry to help developers
monitor, debug, and optimize their agent
workflows. And I guess I saw very few a
few hands up, but how many of you were
collecting traces for eval?
I really Okay, I see a few hands up in
the audience. I see a few hands up. I'm
sorry. I'm like peeking over here. It's
quite bright.
So if you're collecting traces using
open telemetry, you can send those
traces to any destination using Versel
drains. Versel drains allows you to
export all of your Versel data to a
third party observability tool. So if
you're using a third party evals tool,
you can export your traces there.
To make it really easy for developers to
run evals and monitor model quality,
we've actually partnered with Brain
Trust through the Versell marketplace.
With this new integration, you can
automatically stream traces and
evaluation data from Versel to Brain
Trust with just a few clicks. Earlier
today, you may have heard more about our
marketplace integration during a panel
with Anker Goyle, the founder of Brain
Trust.
To make it really easy for you to get
started, we've actually got a demo app
here. You can scan this QR code. Um,
this demo app you can clone from our
template library. It's an AI chatbot
built with Nex.js, the AI SDK and AI
gateway with traces being sent to Brain
Trust. I'll give you all a second.
Oh, am I in the way? Let me get out the
way.
Awesome.
As you can see, our goal is to provide
you with flexibility and control over
what tools you use. While we're working
hard to build really great native
observability tooling, you'll always
have the freedom to send your data
wherever you want. The most important
thing is that you're able to build great
user experiences and turn unpredictable
code into reliable systems.
And with that, that's the end of today.
If you catch me at the happy hour, I'd
love to learn more about what you're
building. Thanks so much.
[Music]
Learn more about how teams are using Vercel's AI native observability tools to build performant and reliable AI applications Get a demo today: https://vercel.com/contact/sales/demo