WEBVTT

00:00:00.000 --> 00:00:01.940
Billions of dollars are currently flowing in

00:00:01.940 --> 00:00:05.719
massive endless circles. Yeah, it is a deeply

00:00:05.719 --> 00:00:08.960
weird financial loop. They do this just to keep

00:00:08.960 --> 00:00:12.240
the machines thinking. Beat. The sheer scale

00:00:12.240 --> 00:00:15.259
of this infrastructure is truly staggering. It

00:00:15.259 --> 00:00:17.460
really is a wild landscape out there. Welcome

00:00:17.460 --> 00:00:20.399
to the deep dive. Today, we are mapping that

00:00:20.399 --> 00:00:23.039
exact ocean for you. We definitely have a lot

00:00:23.039 --> 00:00:24.719
of ground to cover. I was sitting here reflecting

00:00:24.719 --> 00:00:27.350
on this earlier today. The modern artificial

00:00:27.350 --> 00:00:30.370
intelligence race is vast and complex. Honestly,

00:00:30.530 --> 00:00:32.829
I still wrestle with prompt drift myself. Oh,

00:00:32.850 --> 00:00:34.289
absolutely. You know what I mean by that, right?

00:00:34.369 --> 00:00:36.890
You ask an AI for help with a specific task.

00:00:37.390 --> 00:00:41.250
Over time, it slowly starts giving lazier, wildly

00:00:41.250 --> 00:00:43.350
different answers. Getting these complex machines

00:00:43.350 --> 00:00:46.189
to listen is incredibly hard. Yeah. We definitely

00:00:46.189 --> 00:00:48.149
all struggle with that exact same problem. It

00:00:48.149 --> 00:00:50.829
is a universal friction point right now. So here

00:00:50.829 --> 00:00:53.159
is our roadmap for you today. We are exploring

00:00:53.159 --> 00:00:55.500
the massive infrastructure war fueling this.

00:00:55.700 --> 00:00:58.219
We are also looking at a major software shift.

00:00:58.340 --> 00:01:01.020
We're moving away from chatbots toward fully

00:01:01.020 --> 00:01:03.679
autonomous agents. Which is a huge structural

00:01:03.679 --> 00:01:07.200
change. Exactly. And finally, we will explore

00:01:07.200 --> 00:01:10.579
a massive technical breakthrough. Artificial

00:01:10.579 --> 00:01:13.040
intelligence is literally starting to fine -tune

00:01:13.040 --> 00:01:16.420
its own brain. Let's start by looking at the

00:01:16.420 --> 00:01:19.140
Deep Hardware Foundation. Right. Before we analyze

00:01:19.140 --> 00:01:22.079
AI behavior, We must understand the machinery.

00:01:22.659 --> 00:01:24.959
We really need to follow the money loop driving

00:01:24.959 --> 00:01:28.459
everything. It keeps this entire ecosystem alive

00:01:28.459 --> 00:01:31.659
right now. NVIDIA is the absolute center of gravity

00:01:31.659 --> 00:01:35.280
here. Their CEO recently made a highly publicized

00:01:35.280 --> 00:01:39.340
philanthropic move. His foundation donated $108

00:01:39.340 --> 00:01:42.760
million. This was provided entirely in the form

00:01:42.760 --> 00:01:45.260
of computing power. That is a massive operational

00:01:45.260 --> 00:01:47.459
grant. It went straight to universities and nonprofit

00:01:47.459 --> 00:01:49.859
labs. That sounds incredibly generous on the

00:01:49.859 --> 00:01:52.519
surface to anyone listening. Academic researchers

00:01:52.519 --> 00:01:55.180
need that expensive GPU infrastructure desperately

00:01:55.180 --> 00:01:57.299
today. Yeah, they really do. Scientific projects

00:01:57.299 --> 00:01:59.299
require immense amounts of raw computing power.

00:01:59.500 --> 00:02:01.560
They're normally priced entirely out of the modern

00:02:01.560 --> 00:02:03.680
market. It's definitely a huge win for those

00:02:03.680 --> 00:02:05.799
specific universities. But the more interesting

00:02:05.799 --> 00:02:08.259
part is the underlying mechanism. Right. The

00:02:08.259 --> 00:02:10.939
actual logistics of the donation. Exactly. We

00:02:10.939 --> 00:02:12.719
need to look at what this reveals. The foundation

00:02:12.719 --> 00:02:15.699
purchased all of this compute directly from CoreWeave.

00:02:15.800 --> 00:02:18.180
And here is where it gets deeply interesting

00:02:18.180 --> 00:02:21.270
for us. We really have to look closely at the

00:02:21.270 --> 00:02:26.069
timeline. January 2026 was a massive month for

00:02:26.069 --> 00:02:30.289
them. NVIDIA invested $2 billion directly into

00:02:30.289 --> 00:02:32.949
CoreWeave. They didn't just buy shares in a random

00:02:32.949 --> 00:02:35.870
cloud provider. They practically bought the entire

00:02:35.870 --> 00:02:38.590
market's future direction. Right. They became

00:02:38.590 --> 00:02:41.370
the second largest shareholder of CoreWeave overnight.

00:02:41.669 --> 00:02:44.830
And CoreWeave heavily depends on NVIDIA GPUs

00:02:44.830 --> 00:02:47.030
to actually function. There's also a massive

00:02:47.030 --> 00:02:49.849
financial safety net in place here. NVIDIA arranged

00:02:49.849 --> 00:02:53.409
a $6 .3 billion deal. Which is just staggering.

00:02:53.770 --> 00:02:56.169
They will simply buy any capacity CoreWeave feels

00:02:56.169 --> 00:02:58.560
to sell. It's an unprecedented level of corporate

00:02:58.560 --> 00:03:00.620
market protection. Right. And that is a massive

00:03:00.620 --> 00:03:03.039
market guarantee for them. AI demand is growing

00:03:03.039 --> 00:03:05.560
way faster than global compute supply. Yeah.

00:03:05.659 --> 00:03:08.159
This isn't just a simple technology market trend

00:03:08.159 --> 00:03:11.599
anymore. The AI race is becoming a pure, brutal

00:03:11.599 --> 00:03:14.240
infrastructure war. Critics have a very specific

00:03:14.240 --> 00:03:17.400
name for this controversial practice. They call

00:03:17.400 --> 00:03:19.800
this strategy circular financing in the tech

00:03:19.800 --> 00:03:23.340
industry. You invest heavy capital directly into

00:03:23.340 --> 00:03:26.580
your own customers. It bolsters their corporate

00:03:26.580 --> 00:03:29.599
balance sheets significantly and artificially.

00:03:29.680 --> 00:03:32.080
That allows those customers to buy more of your

00:03:32.080 --> 00:03:34.939
product. It is a brilliantly aggressive corporate

00:03:34.939 --> 00:03:37.960
strategy. In this specific case, they are buying

00:03:37.960 --> 00:03:41.250
thousands of GPUs. It's like an automaker giving

00:03:41.250 --> 00:03:44.770
a rental car company cash, but only under the

00:03:44.770 --> 00:03:47.150
strict condition they buy those specific cars.

00:03:47.310 --> 00:03:50.289
Right. Or a casino giving players free chips

00:03:50.289 --> 00:03:52.710
to gamble with, but they own the hotel where

00:03:52.710 --> 00:03:54.689
you eventually spend them. That casino analogy

00:03:54.689 --> 00:03:57.169
is a perfect way to visualize it. It keeps the

00:03:57.169 --> 00:03:59.939
heavy capital flowing in a closed loop. The money

00:03:59.939 --> 00:04:02.500
never really leaves the NVIDIA ecosystem at all.

00:04:02.659 --> 00:04:04.020
We also have to look at the emerging hardware

00:04:04.020 --> 00:04:06.740
competitors. Yeah. We must contrast this NVIDIA

00:04:06.740 --> 00:04:09.080
dominance with the Cerebras IPO. Oh, that was

00:04:09.080 --> 00:04:11.120
an absolutely massive market debut recently.

00:04:11.460 --> 00:04:15.599
Cerebras raised $5 billion right out the gate.

00:04:15.699 --> 00:04:19.959
Wow. Their shares surged 108 % immediately. The

00:04:19.959 --> 00:04:22.420
market clearly wants alternative compute options

00:04:22.420 --> 00:04:26.680
right now. They hit a $66 billion valuation almost

00:04:26.680 --> 00:04:29.279
instantly. Their revenue growth is just absolutely

00:04:29.279 --> 00:04:33.120
staggering to watch today. In 2025, their revenues

00:04:33.120 --> 00:04:37.079
grew 76 % entirely. That is incredible velocity.

00:04:37.379 --> 00:04:41.879
They reached $510 million in total. The sheer

00:04:41.879 --> 00:04:44.660
velocity of that financial growth is stunning.

00:04:44.899 --> 00:04:48.920
We really need to unpack why investors care so

00:04:48.920 --> 00:04:52.180
much. Why is the broader market so desperate

00:04:52.180 --> 00:04:54.660
for an alternative? Because right now, the entire

00:04:54.660 --> 00:04:57.620
tech sector is deeply bottlenecked. The AI on

00:04:57.620 --> 00:04:59.560
your personal phone depends on this hardware.

00:04:59.740 --> 00:05:03.240
If one single company controls all the raw computing

00:05:03.240 --> 00:05:06.439
power, they fundamentally dictate the pace of

00:05:06.439 --> 00:05:09.060
global innovation for everyone. Developers are

00:05:09.060 --> 00:05:11.759
starving for hardware that isn't completely monopolized.

00:05:11.959 --> 00:05:13.899
It makes you wonder about the long -term stability

00:05:13.899 --> 00:05:17.500
here. Is this circular financing a dangerous,

00:05:17.720 --> 00:05:24.800
fragile economy? It's, you know, a fiercely debated

00:05:24.800 --> 00:05:27.019
topic among economists right now. Some see it

00:05:27.019 --> 00:05:29.540
as highly artificial and dangerous market inflation.

00:05:29.879 --> 00:05:32.240
Others argue it's entirely necessary for the

00:05:32.240 --> 00:05:34.860
industry to survive. This new hardware requires

00:05:34.860 --> 00:05:38.139
unprecedented upfront capital to actually exist.

00:05:38.420 --> 00:05:40.600
You have to guarantee the supply chain somehow

00:05:40.600 --> 00:05:44.939
to scale. So they are essentially self -funding

00:05:44.939 --> 00:05:47.800
demand to dominate the entire compute board.

00:05:48.620 --> 00:05:51.040
That is the harsh reality of the hardware game

00:05:51.040 --> 00:05:53.379
today. You either buy the board or you lose the

00:05:53.379 --> 00:05:55.879
game. All that massive compute is enabling a

00:05:55.879 --> 00:05:59.399
profound software shift. We are moving away from

00:05:59.399 --> 00:06:02.360
just chatting with our AI. We're letting AI take

00:06:02.360 --> 00:06:04.759
the steering wheel entirely now. We are officially

00:06:04.759 --> 00:06:07.319
entering the era of the autonomous agent. We

00:06:07.319 --> 00:06:09.339
should define that specific industry term clearly

00:06:09.339 --> 00:06:12.759
first. An agent is software that plans and does

00:06:12.759 --> 00:06:15.889
tasks on its own without supervision. That independence

00:06:15.889 --> 00:06:18.670
is the absolute key distinction to remember here.

00:06:18.790 --> 00:06:21.170
A chat bot waits patiently for your next prompt

00:06:21.170 --> 00:06:24.230
to act. An agent looks at the goal and executes

00:06:24.230 --> 00:06:27.110
multiple steps. Anthropic is pushing this boundary

00:06:27.110 --> 00:06:29.689
aggressively right now. Their cloud code software

00:06:29.689 --> 00:06:31.589
has a brand new goal feature. It's currently

00:06:31.589 --> 00:06:34.029
going completely viral across online developer

00:06:34.029 --> 00:06:36.569
communities. Yeah, I saw that. People are jokingly

00:06:36.569 --> 00:06:39.949
calling it the Ralph Wiggum loop. The Ralph Wiggum

00:06:39.949 --> 00:06:42.939
loop. That is incredibly funny to me. It really

00:06:42.939 --> 00:06:45.019
is. But the underlying mechanism is fascinating.

00:06:45.279 --> 00:06:47.839
It just keeps working autonomously on complex

00:06:47.839 --> 00:06:50.819
coding problems. It loops relentlessly until

00:06:50.819 --> 00:06:54.100
your specific task is fully completed. It figures

00:06:54.100 --> 00:06:55.800
out the errors and corrects itself automatically.

00:06:55.839 --> 00:06:58.660
Exactly. They're expanding into highly professional

00:06:58.660 --> 00:07:01.860
fields, too. Anthropic just released a major

00:07:01.860 --> 00:07:04.899
toolkit called Claude for Legal. It's a completely

00:07:04.899 --> 00:07:08.259
free resource for the legal industry. It turns

00:07:08.259 --> 00:07:11.790
Claude into a highly... customizable AI legal

00:07:11.790 --> 00:07:14.170
assistant. It connects seamlessly with enterprise

00:07:14.170 --> 00:07:16.790
apps like Slack and DocuSign. Yeah. And they

00:07:16.790 --> 00:07:19.509
also gave developers tiny physical computers

00:07:19.509 --> 00:07:21.529
recently. This happened at their latest major

00:07:21.529 --> 00:07:24.129
software developer conference. I was reading

00:07:24.129 --> 00:07:26.009
about those miniature rigs. People instantly

00:07:26.009 --> 00:07:28.430
started building weird, cool hardware projects

00:07:28.430 --> 00:07:30.930
with them. It's sparking a lot of vital grassroots

00:07:30.930 --> 00:07:33.949
developer innovation. Google is also moving into

00:07:33.949 --> 00:07:36.430
this agent space very heavily. Yeah. They just

00:07:36.430 --> 00:07:38.949
launched something called Google Skills to compete

00:07:38.949 --> 00:07:42.839
directly. Right. It offers 13 official pre -built

00:07:42.839 --> 00:07:45.660
AI agent skills. They are fully free and completely

00:07:45.660 --> 00:07:48.399
open source for developers. That is a smart distribution

00:07:48.399 --> 00:07:50.980
play. They work for existing platforms like Clod,

00:07:51.120 --> 00:07:54.199
Cursor, and Copilot. They automate highly advanced

00:07:54.199 --> 00:07:57.459
daily workflows for regular users. There's also

00:07:57.459 --> 00:08:00.259
a major internal leak to discuss here. Something

00:08:00.259 --> 00:08:02.680
called the Gemini Spark agent is apparently coming

00:08:02.680 --> 00:08:05.420
soon. Yeah, the rumors on that are wild right

00:08:05.420 --> 00:08:08.180
now. It will reportedly live natively inside

00:08:08.180 --> 00:08:11.019
the main Gemini app. It may proactively automate

00:08:11.019 --> 00:08:13.860
repetitive work across your various apps. It

00:08:13.860 --> 00:08:16.199
watches what you do and just handles it quietly.

00:08:16.459 --> 00:08:18.579
You won't even need to ask it to perform tasks.

00:08:19.060 --> 00:08:21.160
The creative sector is seeing this exact same

00:08:21.160 --> 00:08:23.699
shift, too. Akesfield just launched a powerful

00:08:23.699 --> 00:08:26.800
agent called Supercomputer recently. It is a

00:08:26.800 --> 00:08:29.779
brand new AI creative agent for designers. It

00:08:29.779 --> 00:08:31.860
plans everything out for you from a basic concept.

00:08:32.080 --> 00:08:34.159
It picks the best models automatically behind

00:08:34.159 --> 00:08:37.000
the scenes. It renders visual assets directly

00:08:37.000 --> 00:08:40.230
from your raw initial idea. The wildest part

00:08:40.230 --> 00:08:42.409
is how the creative process actually starts.

00:08:42.669 --> 00:08:45.389
There is absolutely no prompt needed at all to

00:08:45.389 --> 00:08:47.830
begin. That actually brings up a very serious

00:08:47.830 --> 00:08:50.710
societal concern. I have to push back on this

00:08:50.710 --> 00:08:52.990
concept a bit. Okay, let's hear it. I'm looking

00:08:52.990 --> 00:08:55.269
at creative tools like Hakesfield acting completely

00:08:55.269 --> 00:08:57.929
independently. I'm also looking at consumer tools

00:08:57.929 --> 00:09:01.570
like Alexa for shopping. Alexa now auto -buys

00:09:01.570 --> 00:09:05.149
basic household items for you entirely unprompted.

00:09:05.230 --> 00:09:07.990
Yeah, that is a huge leap in autonomy. It tracks

00:09:07.990 --> 00:09:10.769
prices and remembers past chats across your devices.

00:09:11.289 --> 00:09:13.950
If it auto -buys without us and creates without

00:09:13.950 --> 00:09:16.730
prompts, when do we lose control of the steering

00:09:16.730 --> 00:09:19.470
wheel? Are we losing the steering wheel or finally

00:09:19.470 --> 00:09:22.350
inventing cruise control? We didn't lose control

00:09:22.350 --> 00:09:24.509
when we stopped churning our own butter. That

00:09:24.509 --> 00:09:26.549
is an interesting way to logically frame it.

00:09:26.610 --> 00:09:29.110
We are simply trading tedious micromanagement

00:09:29.110 --> 00:09:32.269
for high -level macro direction. But a car on

00:09:32.269 --> 00:09:34.409
cruise control still needs an active driver.

00:09:34.830 --> 00:09:37.750
Higgs field is rendering complex assets without

00:09:37.750 --> 00:09:40.889
a user prompt at all. It feels like the machine

00:09:40.889 --> 00:09:44.000
is deciding the ultimate destination. It's definitely

00:09:44.000 --> 00:09:46.139
a profound shift in daily human interaction.

00:09:46.440 --> 00:09:48.639
We're moving from being active creators to being

00:09:48.639 --> 00:09:51.159
passive curators. You set the broad parameters

00:09:51.159 --> 00:09:53.360
and let the underlying system run. Let's look

00:09:53.360 --> 00:09:56.000
at a few more examples of this shift. The Kimi

00:09:56.000 --> 00:09:58.980
WebBridge browser extension is absolutely fascinating.

00:09:59.340 --> 00:10:01.980
It can now interact with websites exactly like

00:10:01.980 --> 00:10:05.100
a human would. Yeah, it can search, scroll, click,

00:10:05.240 --> 00:10:07.840
and type dynamically. It completes full multi

00:10:07.840 --> 00:10:10.159
-step tasks right in your active browser. You

00:10:10.159 --> 00:10:11.879
just watch it navigate the web on your screen.

00:10:12.320 --> 00:10:14.779
GrokBuild is also stepping aggressively into

00:10:14.779 --> 00:10:17.299
this competitive arena. It's an agentic tool

00:10:17.299 --> 00:10:20.080
currently operating in a beta phase. Right. It

00:10:20.080 --> 00:10:23.059
handles complex coding, app building, and general

00:10:23.059 --> 00:10:25.620
workflow automation. OpenAI is pushing mobile

00:10:25.620 --> 00:10:28.279
accessibility really hard right now, too. Codex

00:10:28.279 --> 00:10:31.159
is now officially live in the ChatGPT mobile

00:10:31.159 --> 00:10:33.720
app. That is a huge update for developers on

00:10:33.720 --> 00:10:35.360
the go. You can stay connected to active coding

00:10:35.360 --> 00:10:37.740
work from anywhere. It's available on both the

00:10:37.740 --> 00:10:40.940
iOS and Android premium plans. But OpenAI is

00:10:40.940 --> 00:10:43.460
also facing major corporate distribution drama.

00:10:44.029 --> 00:10:47.230
They deeply trusted Apple to push ChatGPT to

00:10:47.230 --> 00:10:50.169
billions globally. They desperately needed that

00:10:50.169 --> 00:10:52.769
massive user base to scale efficiently. The Apple

00:10:52.769 --> 00:10:55.110
partnership reportedly under -delivered very

00:10:55.110 --> 00:10:58.029
badly for them. The integration wasn't as deep

00:10:58.029 --> 00:11:00.970
or ubiquitous as originally promised. OpenAI

00:11:00.970 --> 00:11:03.830
is apparently considering major legal action

00:11:03.830 --> 00:11:06.710
over the failure. Sometimes these complex autonomous

00:11:06.710 --> 00:11:09.409
systems just fail completely on us. They get

00:11:09.409 --> 00:11:12.490
confused by edge cases or highly nuanced requests.

00:11:12.850 --> 00:11:15.389
That is exactly where a platform called Tendom

00:11:15.389 --> 00:11:18.350
comes in. Right. Tendom is an AI platform specifically

00:11:18.350 --> 00:11:21.409
designed for workflow handoffs. Exactly. It acts

00:11:21.409 --> 00:11:25.090
as the ultimate necessary human safety net. You

00:11:25.090 --> 00:11:27.950
submit a complex task in normal, plain language.

00:11:28.230 --> 00:11:31.190
The AI agents try to handle the massive processing

00:11:31.190 --> 00:11:34.470
volume. If they fail, it immediately hands the

00:11:34.470 --> 00:11:37.409
task to a human expert. I want to deeply understand

00:11:37.409 --> 00:11:39.870
the broader impact of this. Which of these tools

00:11:39.870 --> 00:11:41.889
actually changes the fundamental nature of our

00:11:41.889 --> 00:11:44.370
daily work? It's the broader shift away from

00:11:44.370 --> 00:11:47.129
active, hands -on management. We're moving from

00:11:47.129 --> 00:11:50.029
complex prompting to pure outcome -based delegating.

00:11:50.210 --> 00:11:52.690
You're no longer managing the tedious microsteps

00:11:52.690 --> 00:11:54.850
of a task. You're managing the final, polished

00:11:54.850 --> 00:11:57.649
outcomes instead of the process. It's less about

00:11:57.649 --> 00:11:59.789
chatting and more about handing off our entire

00:11:59.789 --> 00:12:03.769
mental load. Two sec silence. Sponsor read inserted

00:12:03.769 --> 00:12:06.870
here. We've seen the incredible billion dollar

00:12:06.870 --> 00:12:10.309
hardware financing loop. We've seen the autonomous

00:12:10.309 --> 00:12:13.750
agent loop automating our software tasks. Now

00:12:13.750 --> 00:12:15.990
we have to ask a much deeper technical question.

00:12:16.309 --> 00:12:19.129
How do we make these autonomous agents truly

00:12:19.129 --> 00:12:22.740
reliably precise? They need to handle high stakes

00:12:22.740 --> 00:12:25.600
enterprise tasks perfectly and safely. Yeah.

00:12:25.720 --> 00:12:27.940
The margin for error is basically zero there.

00:12:28.039 --> 00:12:31.299
I'm talking about parsing highly complex, sensitive

00:12:31.299 --> 00:12:34.759
medical data. or synthesizing highly nuanced

00:12:34.759 --> 00:12:37.419
binding corporate legal documents. That is where

00:12:37.419 --> 00:12:39.419
the latest technical breakthrough changes everything.

00:12:39.659 --> 00:12:41.799
Adaption has just introduced a powerful system

00:12:41.799 --> 00:12:44.580
called Autoscientist. It automates the highly

00:12:44.580 --> 00:12:47.679
complex fine -tuning of AI models entirely. It

00:12:47.679 --> 00:12:49.620
prepares them for very specific high -stakes

00:12:49.620 --> 00:12:52.100
industrial tasks automatically. Fine -tuning

00:12:52.100 --> 00:12:54.500
is usually a highly manual, incredibly tedious

00:12:54.500 --> 00:12:57.259
process. A standard base model is like a smart

00:12:57.259 --> 00:12:59.539
high school student. Fine -tuning is what tunes

00:12:59.539 --> 00:13:01.480
them into an expert corporate lawyer. That's

00:13:01.480 --> 00:13:03.759
a great way to put it. But AutoScientist doesn't

00:13:03.759 --> 00:13:07.220
just train a basic model once. It actively experiments

00:13:07.220 --> 00:13:10.539
on the model continuously over time. It searches

00:13:10.539 --> 00:13:13.299
relentlessly until it finds the absolute perfect

00:13:13.299 --> 00:13:16.190
recipe. The underlying methodology here is really

00:13:16.190 --> 00:13:18.950
quite brilliant to examine. It uses a specific

00:13:18.950 --> 00:13:22.190
technique called dynamic smart selection. It

00:13:22.190 --> 00:13:25.149
automatically tests thousands of different mixtures

00:13:25.149 --> 00:13:28.090
of training data. It evaluates what the model

00:13:28.090 --> 00:13:30.649
actually learns from most effectively. It also

00:13:30.649 --> 00:13:33.450
performs continuous hyperparameter tuning during

00:13:33.450 --> 00:13:36.110
this process. Let's define that dense jargon

00:13:36.110 --> 00:13:38.789
for absolute clarity right now. It means adjusting

00:13:38.789 --> 00:13:41.129
the hidden dials that control how fast an AI

00:13:41.129 --> 00:13:43.690
learns. Think of a human researcher turning three

00:13:43.690 --> 00:13:46.289
dials manually. It takes them days to find the

00:13:46.289 --> 00:13:48.909
exact right combination. Right. Autoscientists

00:13:48.909 --> 00:13:51.769
can spin a million microscopic dials simultaneously

00:13:51.769 --> 00:13:55.730
in seconds. It constantly adjusts those incredibly

00:13:55.730 --> 00:13:58.929
complex internal mathematical settings. It tests

00:13:58.929 --> 00:14:01.590
combinations a human brain couldn't even conceptualize.

00:14:01.750 --> 00:14:04.070
The actual recorded results of this system are

00:14:04.070 --> 00:14:06.870
simply staggering. In rigorous internal tests,

00:14:07.129 --> 00:14:09.409
it didn't just match human experts. Yeah, the

00:14:09.409 --> 00:14:12.129
numbers are wild. It beat highly trained human

00:14:12.129 --> 00:14:15.809
experts by an average of 35%. Base success rates

00:14:15.809 --> 00:14:20.549
jumped from 48 % to 64%. That is a massive operational

00:14:20.549 --> 00:14:24.820
leap. Whoa. Beat. Imagine scaling to a billion

00:14:24.820 --> 00:14:28.159
queries when AI is perfectly tuning itself. It's

00:14:28.159 --> 00:14:31.080
a truly profound moment of technological acceleration

00:14:31.080 --> 00:14:34.200
for us. What makes it even more powerful is its

00:14:34.200 --> 00:14:36.799
broad flexibility. The auto scientist system

00:14:36.799 --> 00:14:40.919
is completely, entirely domain agnostic. It handles

00:14:40.919 --> 00:14:43.919
complex, nuanced legal jargon easily and accurately.

00:14:44.179 --> 00:14:46.840
It synthesizes dense, confusing medical data

00:14:46.840 --> 00:14:49.200
without any hallucination problems. Right. It

00:14:49.200 --> 00:14:52.200
parses complicated, deeply layered finance. reports

00:14:52.200 --> 00:14:55.279
seamlessly for analysts. It proved it could handle

00:14:55.279 --> 00:14:58.120
diverse industries with absolute ease. This solves

00:14:58.120 --> 00:15:00.740
a major persistent problem with modern base models.

00:15:01.019 --> 00:15:03.519
Standard models like Gemini or Claude are incredibly

00:15:03.519 --> 00:15:06.200
versatile tools, but they often lack crucial,

00:15:06.320 --> 00:15:09.039
highly granular enterprise precision natively.

00:15:09.059 --> 00:15:11.500
They struggle with high stakes, specialized enterprise

00:15:11.500 --> 00:15:14.120
tasks normally. They usually require intense,

00:15:14.340 --> 00:15:17.000
expensive human expert tuning to function safely.

00:15:18.800 --> 00:15:21.500
definitively proves that the complex technical

00:15:21.500 --> 00:15:24.340
barrier is falling. You don't need a team of

00:15:24.340 --> 00:15:26.860
PhDs to build a custom model anymore. Exactly.

00:15:26.980 --> 00:15:29.120
I have to ask the logical, slightly terrifying

00:15:29.120 --> 00:15:32.600
next question. If autoscientists consistently

00:15:32.600 --> 00:15:36.740
beats humans at tuning complex AI systems, does

00:15:36.740 --> 00:15:39.919
the human AI researcher eventually become completely

00:15:39.919 --> 00:15:43.519
entirely obsolete? Obsolete is probably too strong

00:15:43.519 --> 00:15:45.899
of a word to use here, but their fundamental

00:15:45.899 --> 00:15:48.899
daily role is definitely shifting quite dramatically.

00:15:49.690 --> 00:15:52.470
Humans move away from manual, tedious mathematical

00:15:52.470 --> 00:15:55.629
dial -turning tasks. They move toward higher

00:15:55.629 --> 00:15:58.169
-level system direction and broad corporate strategy.

00:15:58.549 --> 00:16:01.970
The AI simply handles the deeply complex underlying

00:16:01.970 --> 00:16:05.129
mathematical optimization. The barrier just collapsed.

00:16:05.309 --> 00:16:07.129
The artificial intelligence is literally building

00:16:07.129 --> 00:16:10.210
itself now. It is the ultimate rapidly accelerating

00:16:10.210 --> 00:16:12.889
recursive loop in technology. Let's zoom out

00:16:12.889 --> 00:16:15.029
and look at the whole massive picture. We have

00:16:15.029 --> 00:16:17.429
covered a lot of incredibly dense ground today.

00:16:17.549 --> 00:16:19.909
We really have. The narrative through line of

00:16:19.909 --> 00:16:23.070
this entire ecosystem is deeply connected. It

00:16:23.070 --> 00:16:25.730
starts with billions in circular closed loop

00:16:25.730 --> 00:16:28.929
compute financing. Companies like Nvidia are

00:16:28.929 --> 00:16:31.309
aggressively securing the massive hardware baseline.

00:16:31.509 --> 00:16:33.950
Then that raw compute flows directly into the

00:16:33.950 --> 00:16:36.450
upper software layer. We see fully autonomous

00:16:36.450 --> 00:16:40.240
agents looping digital tasks endlessly. Innovative

00:16:40.240 --> 00:16:42.860
tools like Cloud Code and Gemini Spark take direct

00:16:42.860 --> 00:16:45.480
action. And it all ends with the complex systems

00:16:45.480 --> 00:16:48.980
refining themselves. We have AI systems actively

00:16:48.980 --> 00:16:51.200
fine -tuning their own digital brains. Right.

00:16:51.480 --> 00:16:53.740
Autoscientists is doing it significantly better

00:16:53.740 --> 00:16:56.580
than trained human experts. The entire loop of

00:16:56.580 --> 00:16:58.980
AI development is becoming entirely self -contained.

00:16:59.159 --> 00:17:01.779
It's building incredible momentum at a, frankly,

00:17:01.879 --> 00:17:04.440
terrifying speed globally. The entire tech stack

00:17:04.440 --> 00:17:07.420
is automating its own underlying creation. I

00:17:07.420 --> 00:17:08.680
want to leave you with a final thought today.

00:17:09.079 --> 00:17:11.279
We are watching a fundamental shift in our relationship

00:17:11.279 --> 00:17:14.819
with technology. If raw AI infrastructure essentially

00:17:14.819 --> 00:17:17.640
funds itself entirely today, if agents execute

00:17:17.640 --> 00:17:20.740
our daily corporate tasks entirely autonomously

00:17:20.740 --> 00:17:23.420
tomorrow, and if software systems tune the underlying

00:17:23.420 --> 00:17:26.900
models better than humans can, what exactly is

00:17:26.900 --> 00:17:29.359
the fundamental role of the human in the loop

00:17:29.359 --> 00:17:32.099
tomorrow? As these autonomous systems begin to

00:17:32.099 --> 00:17:34.779
build, fund, and refine themselves continuously,

00:17:35.299 --> 00:17:38.400
are we the architects of this new world? or just

00:17:38.400 --> 00:17:41.059
the audience that is undeniably the defining

00:17:41.059 --> 00:17:42.920
question of our generation right now thank you

00:17:42.920 --> 00:17:45.099
for taking this complex fascinating journey with

00:17:45.099 --> 00:17:47.380
us keep questioning these massive systems as

00:17:47.380 --> 00:17:49.579
they evolve around you keep exploring the distant

00:17:49.579 --> 00:17:52.339
edges of this rapidly shifting new frontier we

00:17:52.339 --> 00:17:54.420
will see you next time utero music