From Nano Banana to Capsules
Key Points
- The speaker frames both human minds and large language models (LLMs) as “jagged” intelligences—highly skilled in some areas (e.g., real‑time motor tasks for humans, earnings‑report summarization for the Nano Banana Pro) but weak in others (formal math for humans, children’s alphabet creation for the model).
- Traditional jobs force individuals to fit their uneven strengths into predefined roles, but the evolving capabilities of LLMs like Nano Banana Pro are reshaping that fit by offering new, more complementary skill sets.
- Each incremental improvement in AI across language, image, video, or 3‑D modeling reduces the “jaggedness” of the technology, instantly unlocking a cascade of downstream business opportunities that were previously infeasible.
- The rapid progression—from a June state where infographics, PowerPoint slides, and marketing images were unsolved, to a December where they are “largely solved”—has already spawned novel ventures such as “capsules,” a new storytelling medium made possible by these AI breakthroughs.
Sections
- Jagged Intelligence and Future AI Applications - The speaker uses the “jagged, intelligent surfaces” metaphor to explain how Nano Banana Pro’s breakthrough reshapes AI capabilities and challenges the traditional practice of forcing unique intelligences into predefined job roles.
- Capsules – Image‑Enhanced Storytelling - Capsules are scrollable narratives that blend generative images—powered by Nano Banana Pro’s image capabilities—with text to convey mood, creating lightweight, attention‑grabbing formats for travel logs, pitches, and other stories previously limited to lengthy threads.
- Layered AI Innovation Cycle - The speaker stresses identifying and plugging “jagged gaps” in foundational AI models—like the new Nano Banana Pro—to quickly spawn multiple downstream businesses, unlocking massive value and illustrating the rapid, layered growth driving the AI revolution.
Full Transcript
# From Nano Banana to Capsules **Source:** [https://www.youtube.com/watch?v=CJFmhEzmOZg](https://www.youtube.com/watch?v=CJFmhEzmOZg) **Duration:** 00:09:54 ## Summary - The speaker frames both human minds and large language models (LLMs) as “jagged” intelligences—highly skilled in some areas (e.g., real‑time motor tasks for humans, earnings‑report summarization for the Nano Banana Pro) but weak in others (formal math for humans, children’s alphabet creation for the model). - Traditional jobs force individuals to fit their uneven strengths into predefined roles, but the evolving capabilities of LLMs like Nano Banana Pro are reshaping that fit by offering new, more complementary skill sets. - Each incremental improvement in AI across language, image, video, or 3‑D modeling reduces the “jaggedness” of the technology, instantly unlocking a cascade of downstream business opportunities that were previously infeasible. - The rapid progression—from a June state where infographics, PowerPoint slides, and marketing images were unsolved, to a December where they are “largely solved”—has already spawned novel ventures such as “capsules,” a new storytelling medium made possible by these AI breakthroughs. ## Sections - [00:00:00](https://www.youtube.com/watch?v=CJFmhEzmOZg&t=0s) **Jagged Intelligence and Future AI Applications** - The speaker uses the “jagged, intelligent surfaces” metaphor to explain how Nano Banana Pro’s breakthrough reshapes AI capabilities and challenges the traditional practice of forcing unique intelligences into predefined job roles. - [00:03:39](https://www.youtube.com/watch?v=CJFmhEzmOZg&t=219s) **Capsules – Image‑Enhanced Storytelling** - Capsules are scrollable narratives that blend generative images—powered by Nano Banana Pro’s image capabilities—with text to convey mood, creating lightweight, attention‑grabbing formats for travel logs, pitches, and other stories previously limited to lengthy threads. - [00:06:53](https://www.youtube.com/watch?v=CJFmhEzmOZg&t=413s) **Layered AI Innovation Cycle** - The speaker stresses identifying and plugging “jagged gaps” in foundational AI models—like the new Nano Banana Pro—to quickly spawn multiple downstream businesses, unlocking massive value and illustrating the rapid, layered growth driving the AI revolution. ## Full Transcript
I've been thinking about Nano Banana Pro
and how it's changing how we think about
what's solvable and what's possible with
AI. I don't want to show you Nano Banana
Pro today. I want to show you something
that was made possible because of the
advances Nano Banana Pro made and help
you start to think about the kinds of
applications for intelligence that are
going to be possible in 2026 because we
made this one breakthrough. So, here's
the mental model. LLMs are jagged,
intelligent surfaces, which frankly
people are too. We're very, very good at
some things. My brain can do the
differential equations in real time that
are needed for me to catch a baseball
with my hand or a cricket ball. I can do
that without thinking about it. As long
as I practice, my brain can figure out
which face goes with who with a shocking
degree of accuracy because we evolved to
be face-to-face people. My brain is
actually not as good at formal
mathematics. If I had to sit down with
differential equations and a piece of
paper, I would need some training,
thinking, and experience to do that on
paper without a calculator. The point is
that they're jagged. We all have
different shapes to our jagged
intelligence. And traditionally in
careers, a lot of the art of it has been
taking your unique jagged intelligence
and sticking it into a predefined role.
Sort of like sticking a special
snowflake shape into a square or a round
hole for work and saying, "This is what
you're going to do." It may not quite
fit your strengths, but it's close
enough and you can do the job. So, LLMs
are changing that because it turns out
LLMs are also jagged, but they're jagged
in very different ways than we are. I
did my Nano Banana Pro review and I
discovered Nano Banana Pro can
accurately summarize an entire earnings
report, but if you ask it to make a
children's alphabet, it cannot do that
yet. And so what we need to realize is
that each advance we make along these
critical axes in language, in image, in
video, in three-dimensional modeling and
others unlocks tremendous numbers of
downstream businesses that can build off
of that breakthrough. Basically, every
time the AI gets a little bit less
jagged and starts to solve a piece of
that puzzle a little bit better, we
unlock a tremendous amount of value. And
Nano Banana Pro did that because even if
it can't make a children's alphabet, it
turns out it can make a pretty darn good
PowerPoint slide, it can do a lot of the
image work we need to a degree of dness
that we just call good and we don't have
to touch it when it comes to business
images. And so infographics, PowerPoint
slides, marketing images, largely a
solved problem in December and largely
all of them were unsolved in June.
That's how fast we're moving. And when
you move that fast, you get really cool
new businesses that unlock. And the one
I want to talk about today is called
capsules. They don't know I'm talking
about this. This will be a surprise.
Capsules is a new kind of storytelling
medium that recognizes that we are
visual creatures. That we are not text
creatures. That even though chat GPT has
hit a billion users, we are still
textriven creatures, images creatures at
heart. And it's easier for us to see
than to read. And so despite the success
of Ched GPT, I think Nano Banana Pro and
the imaged driven revolution that will
follow is going to be even bigger. And
one reason why that may be true is shown
in the growth charts that we're seeing
for Nano Banana Pro. Gemini using Nano
Banana Pro is growing even faster to a
billion users than Chad GPT ever did. It
is taking Chad GPT's place as the
fastest growing app to a billion users.
And I think it's largely off of the
image capabilities of Nano Banana Pro.
So what is what is capsules? Capsules is
a storytelling medium that combines text
with generative images that reflect the
mood of the text and that is scrollable.
And so what it feels like is a little
bit like unraveling a parchment that
shows you a moving picture and helps you
get a sense of the mood of the text as
you scan. And it's a very light, easy,
thoughtful way to read. It works for
travel logs. It works for reflective
pitches. It works for a whole host of
stories that would previously have been
locked in long Twitter threads or
previously locked on an age ago, like a
version of Reddit or even back in the
day MySpace. This is a way of telling
stories that pop, that gets your
attention. And yes, I'm going to show
you one. I think it's a really cool
idea. And one of the things that's
interesting about capsules is it is
clear that they are unleashing a new
generative technology in ways that allow
the community to tell stories that were
not previously possible. And so this is
underlining the larger point for me,
which is that we can't see all of the
downstream implications, but when a new
foundational technology piece hits, when
when images are solved the way they were
with Nano Banana Pro, we then get to a
point where a host of downstream
flywheels can get started because we got
that one core piece solved. So without
further ado, I'm going to show you
capsules. This is a capsule. It might
look like just a classic 2010s web page
with a nice scroll effect, but what
you're seeing is that you're actually
generating live and animating an image
that goes with the concept. In this
case, the writer is pitching a new
startup driven by the idea of a canary
trigger. And so, the story is all about
how canary triggers worked during the
Cold War. What it means to have a dead
man switch where you have something
happen if you are absent. how that has
been difficult to do in the age of the
internet due to technical obstacles and
how now with the blockchain and with AI
they can code up and build a true canary
switch that allows journalists to share
information with relevant authorities if
they are ever uh taken hostage or in
other ways interfered with. So this is
something where a startup is basically
saying we built a technology that
enables journalists to be protected and
tell the truth. And this gives us a way
to get that story across that connects
their own personal background in East
Germany during the cold war with the
concept now in the age of the internet
that they're trying to get across. And
so the Canary story just sort of is told
visually. There's this nice animation
effect that's being generated by AI.
They're basically giving us a pitch
deck. And look, it's not my job to tell
you whether this is a good pitch or a
bad pitch. It's my job to say this would
not be possible as a pitch without Nano
Banana Pro a month ago and without a
startup named Capsule that built over
the top of Nano Banana Pro and is now
enabling us to tell these kinds of
stories. That is what is interesting.
And so I don't care really if these guys
get funded or not. What I care about is
that we understand the pattern. When you
get a new foundational technology that
fills in one of the jagged troughs in AI
intelligence, the jagged gaps in AI
intelligence, and it's a little bit
better now, you unlock a tremendous
amount of value. Tens of billions,
hundreds of billions of dollars in value
because all kinds of downstream
businesses like Capsule and like this
new startup building on top of Capsule
immediately spin up. You're already two
generations in on your business lineage.
You have Nano Banana Pro. You have
Capsule built on top of Nano Banana Pro
to tell stories. And now yet a third
business that is using this brand new
technology in a brand new startup to
tell their own brand new story. That is
what transformational change looks like.
That is what it looks like to be in the
middle of the AI revolution. That's how
fast things are moving. Remember, Nano
Banana Pro is barely a month old and
we're already three lineages down. That
is how fast AI startups are moving to
take advantage of these skills. And so
my challenge to you in 2026 is one, you
should be building with Nano Banana Pro,
but but two, look for the other spaces
where LLMs have jagged gaps and look for
what it looks like to know they're
closed and move quickly. I'll give you
some examples. I think robotic
graspiness and robotic coordination is
an area where we're very close to a big
breakthrough. Another example, I think
always on agents are very close to a big
breakthrough. I think continual learning
is close to a big breakthrough. I think
memory is close to a big breakthrough. I
think proactivity is close to a big
breakthrough. Think about those. What
does it look like to know that those are
solved enough to build in those spaces?
And how do you start to think about your
product needing that kind of capability
and get ready to build for it so that
you know that you can build into that
space before other people do in 2026
once whatever the correct product
release is the Nano Banana Pro
equivalent is released because that's
what you're looking for, right? You're
looking for some signal from the model
makers. We've solved continual learning
or memory is solved enough or there's an
always on agent that works enough and
now you can build into that space. And
if you look at the building blocks, you
can start to see when we're close. So
with Nano Banana Pro, you could see that
a lot of the work on images was very,
very close. Realistic images were there.
Text was there. All we needed was
something that put it all together. And
that's what we got with Nano Banana.
That's what I want to encourage you to
look at. Look at the spaces where we
have almost everything but not quite.
and then plan your product roadmap
accordingly because these businesses are
going to unfold fast. If if we're three
lineages deep on Nano Banana Pro in a
month, how many more businesses will we
unlock in 2026? It's really exciting
stuff. So that's Capsule. I hope that
gives you a sense of how quickly AI
businesses move and of how important it
is to jump in once key LLM gaps are
solved by a foundational technology
shift like Nano Banana Pro. Cheers.