Comet Redefines AI Agents with UI
Key Points
- The speaker has been inundated with AI agent pitches but found none truly impactful until discovering Comet, whose effectiveness stems from its superior user interface rather than raw AI capability.
- Unlike other tools such as Zapier or n8n that require heavy effort to define and maintain specific workflows, Comet aims to function as a general‑purpose assistant that automatically handles tasks without the user needing to manage its inner workings.
- Previous attempts at general‑purpose agents, like OpenAI’s Operator and Gemini’s Project Mariner, suffered from clunky, slow, and unintuitive UI designs that hampered adoption, highlighting the critical role of UI in agent success.
- Perplexity’s Comet succeeds by making the assistant “disappear” – it executes work silently in the background, delivering results without exposing the user to the complexities or cost of building and controlling the agent.
- The speaker emphasizes that a well‑designed UI can turn an AI agent from a novelty into a practical, day‑to‑day productivity tool, and Comet exemplifies this shift.
Sections
- UI‑Driven AI Agent Success - The speaker argues that, unlike other AI agents requiring heavy setup, Comet’s intuitive interface lets the assistant operate autonomously and seamlessly, making it the only agent that genuinely improves their workday.
- Seamless Sidebar Automation with Comet - The speaker highlights how Comet’s sidebar lets them edit approvals, send emails, manage calendars and LinkedIn, and discover local restaurants instantly—emphasizing speed, trust, and an unobtrusive, efficient UI.
- Comet AI Identifies TikTok Layoffs - The speaker praises Comet’s coherent analysis of Amazon’s strategy and demonstrates its capability to automatically pull LinkedIn profiles of TikTok Shop product managers and engineers possibly impacted by recent layoffs, all while integrating with calendar and email tools.
- Rapid Rise of General-Purpose Agent - The speaker highlights an unexpectedly fast, efficient AI agent that seamlessly handles diverse tasks across multiple platforms, suggesting it may outpace the anticipated rollout timeline.
Full Transcript
# Comet Redefines AI Agents with UI **Source:** [https://www.youtube.com/watch?v=rkQhpiLn8EI](https://www.youtube.com/watch?v=rkQhpiLn8EI) **Duration:** 00:11:49 ## Summary - The speaker has been inundated with AI agent pitches but found none truly impactful until discovering Comet, whose effectiveness stems from its superior user interface rather than raw AI capability. - Unlike other tools such as Zapier or n8n that require heavy effort to define and maintain specific workflows, Comet aims to function as a general‑purpose assistant that automatically handles tasks without the user needing to manage its inner workings. - Previous attempts at general‑purpose agents, like OpenAI’s Operator and Gemini’s Project Mariner, suffered from clunky, slow, and unintuitive UI designs that hampered adoption, highlighting the critical role of UI in agent success. - Perplexity’s Comet succeeds by making the assistant “disappear” – it executes work silently in the background, delivering results without exposing the user to the complexities or cost of building and controlling the agent. - The speaker emphasizes that a well‑designed UI can turn an AI agent from a novelty into a practical, day‑to‑day productivity tool, and Comet exemplifies this shift. ## Sections - [00:00:00](https://www.youtube.com/watch?v=rkQhpiLn8EI&t=0s) **UI‑Driven AI Agent Success** - The speaker argues that, unlike other AI agents requiring heavy setup, Comet’s intuitive interface lets the assistant operate autonomously and seamlessly, making it the only agent that genuinely improves their workday. - [00:03:24](https://www.youtube.com/watch?v=rkQhpiLn8EI&t=204s) **Seamless Sidebar Automation with Comet** - The speaker highlights how Comet’s sidebar lets them edit approvals, send emails, manage calendars and LinkedIn, and discover local restaurants instantly—emphasizing speed, trust, and an unobtrusive, efficient UI. - [00:06:27](https://www.youtube.com/watch?v=rkQhpiLn8EI&t=387s) **Comet AI Identifies TikTok Layoffs** - The speaker praises Comet’s coherent analysis of Amazon’s strategy and demonstrates its capability to automatically pull LinkedIn profiles of TikTok Shop product managers and engineers possibly impacted by recent layoffs, all while integrating with calendar and email tools. - [00:10:53](https://www.youtube.com/watch?v=rkQhpiLn8EI&t=653s) **Rapid Rise of General-Purpose Agent** - The speaker highlights an unexpectedly fast, efficient AI agent that seamlessly handles diverse tasks across multiple platforms, suggesting it may outpace the anticipated rollout timeline. ## Full Transcript
My inbox is littered with people
pitching me AI agents. If I go online, I
see nothing but AI agents. It's the year
of the AI agent, and I have yet to find
an AI agent that really made a
difference in my workday until today.
Comet is that AI agent. And the reason
why is not AI, it's UI, it's user
interface. And I'm going to get into why
that is and kind of show off Comet a
little bit. at the end of the day is an
agent that works because the team at
Perplexity has figured out that my goal
is to actually let the assistant do the
work and disappear. It is not to see
everything the assistant does. It is not
to get buried in the cost of building
the assistant. And so many agents right
now bury you in the cost of building the
assistant or controlling it directly.
Let me give you a few examples. Zapier
and NA very different generation. Zapier
is a much older company. and it ends
much newer. Both of them lean you into
building an agent. You have to define
exactly what you want it to do. You have
to invest really, really heavily in
getting it to work right. N8 is soaring
in popularity because it is better at
that. It does offer value. Don't hear me
wrong, but it is not a general purpose
agent. It's not close to a general
purpose agent. And if you try to build a
general purpose agent in NA, you're
probably going to be disappointed. It is
useful for specific agentic tasks. If
you want to do a specific job that
involves extracting text from a document
and putting it into a spreadsheet, for
example, Naiden will do fine at that.
You can probably get Zapier to do fine
at that. There are a lot of other agent
startups that will do that, too. But
Comet is different. Comet by Perplexity
is really a generalurpose agent
assistant. The closest comparison we've
had previously has been Operator from
OpenAI and Project Mariner, which most
people don't know about from Gemini.
Look, I wanted to love Operator. I
really did. operator. The vision was
great. Chad GPT on the web, the power of
Chad GPT decisioning across the web. The
execution and the UI were really, really
bad. They just were, I'm not going to
sugarcoat it. It is awkward to have this
tiny little browser that looks like a
toy-sized browser inside a chat window.
It's slow. Like, I don't care what the
actual clock time says on these Asian
things. I think it starts the clock late
or something because I swear it takes
longer than it says it takes. I've had
instances where operator says it took
eight minutes and it was like 20 minutes
of clock time. And so if you think about
it, the vision for operator was great,
but what doomed it was the UI. It was
the UI. And what Perplexity gets right
is the UI. That's why Comet shines.
Comet shines because the fundamental
inside of the Perplexity team is that
the assistant should disappear. They
should just go do work for you. I was on
a different website today. I was
actually looking at Substack cuz of
course I was and I was just like crap I
have a meeting I have to reschedule it
and so I just worked with Perplexity to
get the meeting rescheduled. I talked to
my common assistant and I said hey this
is the meeting approximately here is
where I want it moved. There's a block
of meetings here. Suggest a time. It
went it found the meeting. It suggested
the time. It actually gave me a
recommendation for a better time within
that block that would fit well with my
goal of like clean working blocks. And
then it drafted up um a calendar change.
I made a minor edit and approved it. And
then it said, "Hey, just as a best
practice, you should send an email to
this person, too." And I said, "Okay,
great. I'll send an email." I drafted up
the email and I made another minor tweak
and approved it and sent it all from the
sidebar while I was doing other things.
I didn't have to go to Gmail. I didn't
have to go to Gcal. It was super easy.
LinkedIn. I was able to go through and
look at my pending LinkedIn invites
sitting in the sidebar chatting with
Comet. Super super easy. What I'm saying
is that the idea that we have to see the
agent is probably a legacy of the idea
that the agent is untrustworthy. And if
the agent can hook into data and it just
works, we don't need to supervise it as
closely and we kind of don't want to. We
want it to be fast and trustworthy and
dependable. And I will say comet gets
this stuff done quickly. It gets it done
really well. I am a fan, as you might
expect, of Indonesian cuisine. It found
for me an Indonesian restaurant that I
had not run across in my neighborhood.
Well, not quite my neighborhood, two or
three neighborhoods over, but still, it
found something in a special interest of
mine in my city that I wouldn't have
been able to find otherwise. So, I'm
getting real value on day one, and it's
been really easy. I also think when
Comet needs to interact with your
browser and starts to take over, the way
they've done that UI is really good.
What Comet does is it actually goes to
your main browser and glows it up blue
like you're like looking through a Star
Trek portal or something and then it
like does something quickly in the
browser and then like it takes the blue
away. It's a very intuitive way of
saying, "Hey, Comet is driving now. You
always have a chance to stop it." It's
just really smoothly implemented. You're
probably tired of hearing me talk about
Comet. So, let's very quickly show you
Comet. Okay, here we are. Tik Tok shop
is slowing down. That is I literally put
that into the address bar here. It's
triggers a perplexity search. It talks
about the impact of how Tik Tok shop is
slowing down. I have a thesis that I
want to explore with assistant and we're
just going to watch Assistant work.
Okay, I have a thesis
that Tik Tok shop slowing down affects
social commerce more broadly and may
impact
Amazon's outlook.
Please look at Amazon's
recent public info
filings, etc. Get a sense of how the
company is looking at social commerce.
And you can see it's just working away
there. Now if I want to continue, I can
also continue here. I can say
has Tik Tok experienced layoffs around
Tik Tok shop right this is happening
simultaneously. So Amazon's produ
pursuing an agnostic social commerce
approach explicitly response to Tik Tok.
uh similar integrations with Instagram,
Snap, Pinterest, and YouTube Shorts.
Growing
a bunch of other initiatives to keep up
with creatives.
So, I'm getting a sense of how Amazon's
approaching this,
and I really appreciate this is actually
a more coherent take than I typically
see from AI at this point.
It's it's not just bland like Amazon is
not ignoring tech stock slowdown. It's
hedging risk and investing further. That
seems to be supported by the comment up
there. And it's also more actionable and
on point than I've typically seen. Uh
and yes, this has affected Tik Tok shop.
So, it seems like we have a Tik Tok shop
impact. All right, let's just do
something fun. Can you find the uh top
profiles on LinkedIn for Tik Tok shop
affiliated
PMs and engineers
who may have been affected by layoffs
and it's going to go and like look at my
LinkedIn. So what what I'm trying to
give you a sense of is how this can look
across multiple things. It plugs into
LinkedIn. It plugs into your calendar.
It plugs into your Gmail. You can do
multiple things at once. This is going
to keep running even if I change where
I'm going up here in the address bar
and then it starts to find them and I
don't know like I'm not going to go and
click on their links. I don't want to
violate their privacy. But like it gives
me both a way to approach it and some
example profiles. Overall it's pretty
good.
So there you go. That's Comet. I have
been really pleased with it. I think
that Comet is the first agentic browser
that actually lives up to the name. And
I think whether it's worth $200 a month
essentially requires you to add up those
5, 8, 10, 15 minute increments that it's
going to be saving you and be
disciplined about it, measure the value
of your time and say, is this worth $200
a month in time savings to me? And
that's a new way of valuing software.
But I think that's where we're at with
cognitive uh intelligence baked into
software at this point. It gives us a
new valuation paradigm for software that
I'm still getting used to myself. I'm
not used to paying $200 a month for any
kind of software, let alone for a
browser. Old Nate is crying. I don't
know why I'm doing this. But it makes
sense when you think about the time
savings. If this thing saves me 10 15
hours a month, it's obviously worth the
$200. And if you add up all those 10 15
minute increments saved through the day,
I can see the case for 10 hours a month
in savings and that would make it worth
it. So I think my challenge to you if
you want to start to explore good agents
is to ask yourself to impose the kinds
of tests you would actually use them
for. Be really rigorously honest with
whether they work or not. Like I put
Comet through eight different workflow
tests this morning just to see how good
it is before making this video. I was
not easily impressed. I wanted to
examine it and I wanted to examine it
versus operator because operator is a
tool I know pretty well and been frankly
disappointed in. If you want to try
Comet, I would suggest signing up for it
for just a month and putting it through
workflow test in your own environment
with the tools you use and see if you
get value and measure the value. Measure
the time savings. If you're going to
make an investment like this, make sure
you make it worth it. Because if it is
worth it, if it saves you more than 200
bucks a month, it's going to be probably
a multiple of that. Like, if it's saving
you 10 hours and you value your time at
$35 an hour, you're getting close to 2x
ROI every single month. If you value
your time more highly, you get more. If
you save more time, you get more. You
see where the math goes. And that
doesn't count it helping you make better
decisions. And because it can look at
the browser natively, it can help you
make better decisions in a way that like
screenshotting doesn't really do. And so
there's another sort of hard to measure
aspect where there's something native
about the collaboration between the
assistant and the browser that you can't
get any other place cleanly. And I think
that's another piece of value that
perplexity is positioned to capture. Now
the key if you're perplexity is that you
are making an OS, right? This is where
all of these AI players are going. We
live on the web so much that if you
become the browser, the dominant browser
of choice, you become the OS for AI. And
that's where they're going. Time will
tell if this actually pays off. If
Google launches Mariner more
aggressively because Google obviously
has a stake in the ground with Chrome.
Maybe Open AI is going to lean away from
voice a little bit and lean farther into
operator. Who knows? One thing we know
about AI is that it's going to change
frequently. But for now, what I know is
that I actually saw a general purpose
agent that work today. And I'm still
amazed. I never expected to see a
general purpose agent that was this
fast, this efficient, this effective
across a wide range of tasks, flight
reservations, restaurants, looking at
LinkedIn tasks, looking at Gmail tasks,
looking at calendar tasks, looking
across multiple different sources to
accomplish workflows. It's doing a lot.
And I can see that the team has built a
data structure that's going to enable
them to do even more. To me, this feels
ahead of schedule. Yes, I knew agents
were coming, obviously. Yes, I knew
general purpose agents are coming. This
fast and this connected feels three or
four months early. So, it's great to see
it. And we'll see sort of where the
Perplexian Comet team takes us next. Let
me know if you get comment what you
think of