A Wall Street Genius's Final Investment Playbook-Chapter 286: The Invisible Hand (21)

Meanwhile, as the AI power struggle between the U.S. and China intensified and the AI boom set the market ablaze, Gooble’s internal atmosphere was heavier and more complicated than ever.

“The cloud business is booming. If we go by the financials alone, the performance is flawless.”

Gooble’s AI infrastructure business, built on its cloud platform, continued to grow impressively. However, Gooble couldn’t simply celebrate the situation.

“The problem is… we’re gradually losing our presence in the market.”

That’s because at the center of this AI frenzy were not Gooble’s Reinforcement Learning (RL) models—but Next AI and its LLM (Large Language Model) approach.

Gooble’s once-proud RL tech had now become a forgotten relic of the past.

“It’s ironic, isn’t it? The AI era we longed for has finally arrived.”

“And yet, how did we end up being pushed out like this…”

Just a few months ago, things were completely different. Back then, there was a clear rivalry: “Stark vs. Gooble,” “LLM vs. RL”—and Gooble was undeniably one of the two AI giants driving the debate.

But all of that was overturned in an instant—because of one rumor:

“China is trying to recruit Next AI.”

That issue, when merged with the explosive backdrop of a presidential election, quickly shifted the discussion away from technological merit to a narrative of national security and tech protectionism.

“There’s only one reason people are backing LLMs right now—fear. The fear of ‘never letting it fall into China’s hands.’”

This trend wasn’t driven by the true value of technology, but by an emotional impulse—a sense of urgency to protect, rather than rational comparison.

“In this climate… no matter how much we improve our products or technology, it won’t matter. Because the difference didn’t arise from logic to begin with.”

In the end, after much deliberation, Gooble made a decision:

“We need to change direction too. Toward an LLM-centered strategy.”

At first, Gooble’s leadership was shocked and confused… But they soon regained their composure and ran the numbers—and they arrived at a rather favorable conclusion.

‘In this case… the situation actually plays to our advantage!’

If the competition came down to hybrid models that combined RL (Reinforcement Learning) and LLM (Large Language Models), then the winner would ultimately be determined by technical superiority. The key was how well-rounded a team was in both areas.

And since Next AI had already open-sourced their LLM code, there were essentially no barriers for Gooble to catch up on the LLM side.

On the other hand, RL?

Gooble had spent years and massive resources to build unrivaled expertise in reinforcement learning. Whereas Next AI had almost zero experience or infrastructure in that area.

In a hybrid-model face-off, Gooble had a massive upper hand.

As Gooble’s executives quietly celebrated this promising outlook—Ha Siheon spoke again. And what he said next took them by surprise:

“So, I propose a partnership. Next AI brings its expertise in LLM, and Gooble provides its strengths in RL.”

“……”

“……”

A brief silence followed. Gooble’s top brass simply stared at Ha Siheon in a daze.

‘Did I hear that right?’

Ha Siheon was now asking Gooble to share the RL expertise they’d painstakingly built over years—and he spoke as if he were doing them a favor.

‘Why should we?’

‘Did we miss something?’

No matter how they looked at it—nothing added up. Breaking the awkward silence, someone finally spoke.

“But Next AI’s LLM is open source—it’s already freely available to anyone, isn’t it?”

“Yes, that’s correct.”

“……?”

So basically… Ha Siheon was offering a free resource—and in return, asking for Gooble’s exclusive, proprietary tech. One of the executives asked in disbelief:

“Isn’t it unreasonable to expect us to agree to such a partnership?”

“Ah, a reason? Of course, there is one. We need RL technology. And Gooble happens to be the leader in that field. We saw no need to look elsewhere.”

“……??”

It was a completely one-sided line of logic.

“No, no—I mean, sure, it makes sense for Next AI, but from Gooble’s perspective, there’s no reason to share our technology with a competitor.”

“Ah, that’s the reasoning you meant?”

Ha Siheon chuckled briefly and replied:

“It’s simple. For the future of the United States.”

“Excuse me?”

“You’re all well aware of the current global situation, aren’t you?”

“……”

At that moment, Gooble’s executive team went stiff. They were finally starting to grasp Ha Siheon’s real intent.

“With China eyeing global AI supremacy, should American companies really be undercutting each other? I believe now is the time to unite, not squabble over profits.”

Ha Siheon’s argument wasn’t based on logic or technical data. His pitch was built on something else entirely:

Patriotism.

“Of course, we’ll understand if Gooble declines. But in that case, we’ll have no choice but to seek a partner in another country—one with strong RL capabilities. And if that happens… Then this technology will no longer be America’s alone. And that worries me.”

In essence, he was saying: “Partner with us—or we’ll team up with foreign firms.”

And though politely phrased, it sounded eerily like a threat.

Now that LLMs had become the undisputed industry standard, it was unwise to resist the trend. So Gooble made the tough decision to pivot its product strategy toward LLMs.

“No matter what anyone says, the leader in LLMs right now is undoubtedly Next AI. That makes us a complete latecomer in the game.”

In any market, the first mover has a huge advantage—brand recognition, market share, development speed… They lead on every front. In such an uneven match, there was only one path for a latecomer to catch up:

“We need differentiation. We have to fully leverage the one technology where we undeniably have the upper hand.”

Fortunately, Gooble had a trump card it could play.

“Let’s use RL. We’ll front a hybrid model that combines traditional LLMs with reinforcement learning.”

What they unveiled was an integrated technology strategy—a vision to leap beyond Next AI’s current models by offering something more advanced. By doing this, Gooble could ride the LLM wave, while still capitalizing on its own unique strengths.

But then—

“Ha Siheon has requested a meeting.”

“……”

“……?”

“……”

The mood in the meeting room instantly froze.

“What does he want?”

“He says… he’d like to propose ‘reconciliation and cooperation.’”

“Reconciliation and cooperation…?”

Everyone had the same thought flash through their minds: ‘Now, of all times?’

Still, they couldn’t reject the meeting outright.

“If we refuse to meet him…”

“That could be a problem. If we reject his peace offer, he’ll immediately treat us as an enemy!”

Collaborating with Ha Siheon was extremely uncomfortable, but ignoring him and risking turning him into a full-blown adversary? That felt far worse.

“Let’s… hear him out, at least.”

With the meeting confirmed, Gooble’s leadership held an emergency strategy session. There was only one agenda item:

“What does he really mean by ‘cooperation’?”

“At this point, there’s no reason he should need us…”

Ha Siheon was clearly the winner of the AI war with Gooble. He now had full-blown government support. Why, then, would he be the one to reach out?

Many theories were thrown around, but none made sense—until a few days later, when Ha Siheon showed up in person and revealed his proposal:

“I’d like to launch a joint project with Gooble.”

“A… joint project?”

“Yes. Next AI plans to develop an integrated model that combines LLM with RL.”

In that moment, Gooble’s executives froze. Ha Siheon’s strategy was identical to the one they had painstakingly arrived at themselves.

“A hybrid model…”

“So they were… thinking the same thing?”

“Of course, as a business organization that pursues profit, I respect that decision… But I’m deeply concerned about the potential consequences.”

“……”

“……”

The more they thought about it, the clearer it became: This was a threat.

America was boiling with nationalism under Trenton’s leadership. And in such a climate, if key AI technology were to leak overseas?

“Once again, I assure you, we sincerely hope this technology remains solely in American hands. And we truly hope Gooble shares that same commitment.”

If the partnership failed, the blame would be squarely placed on Gooble. It wasn’t hard to imagine headlines across the country proclaiming: "Because of profit-driven Gooble, America’s AI technology was leaked abroad."

And this man in front of them—had already proven more than once that he could inflame public sentiment on national TV.

“If we reject this…”

Even releasing a good product wouldn’t save them. Gooble would be branded as a “traitorous company” overnight.

Ha Siheon delivered his final words with a gentle smile:

“We will fully respect whatever decision Gooble makes. The choice is yours.”

“……”

“……”

“……”

“We’ll review it and get back to you.”

Gooble wrapped up the meeting with only a noncommittal response. Just before leaving the room, I added a final push:

“This is a rare chance to become national heroes. I sincerely hope you make the wise choice.”

“……”

“……”

“……”

The expressions of those leaving the conference room weren’t exactly cheerful—but I wasn’t particularly worried.

“They’re not stupid. They’ll make the right decision eventually.”

In truth, there wasn’t much to debate. Better to accept my offer than be publicly branded as traitors.

“And I do plan on offering some lip service.”

I have at least a shred of conscience. I know exactly how much time and money went into their RL work, and I had no intention of taking it for free.

For example, I’d make sure to publicly say Gooble was playing a vital role in America’s AI future. That kind of recognition would help sustain the partnership.

“Besides, it’s in my best interest to have a hybrid model.”

Developing a treatment for Castleman disease using AI required Gooble’s help. LLMs alone couldn’t overcome certain critical limitations. While LLMs could analyze vast medical data to identify causes of disease or promising drug candidates—they couldn’t determine which of those candidates had the highest chance of success. That’s where RL was essential.

To use a metaphor: If LLMs draw the map, RL finds the fastest and most efficient route on that map.

Anyway—

“Now that this part is settled…”

There was one final thing to do. As I got into the sedan, the driver asked:

“Where to next?”

“Quantum Genome, please.”

Quantum Genome was one of the biotech companies I had invested in—a frontrunner in the field of spatial transcriptomics. It was also the firm I had entrusted with decoding Milo’s biological samples.

Did I mention this before?

Milo’s sample is like a vault. Inside, it holds critical information about Castleman’s “insanity switch.”

“We haven’t been able to unlock it yet…”

Truthfully, the entire AI war I waged was ultimately to acquire the technology needed to open that vault.

There were three keys required to unlock it: GPU, GNN, and Ignus.

I had long since secured the GPU and handed it over. The GNN model—designed to interpret gene interactions and spatial patterns—was delivered last month.

And just last week…

After meticulously optimizing Ignus for parallel computation, memory handling, and compatibility with large-scale spatial transcriptomics data, I passed it along as well.

Now, the result was close at hand.

“It won’t produce results immediately…”

But just as I was trying to manage my impatience, I received a text yesterday:

— We believe we can begin applying the technology to human samples.

The signal that everything was ready. In other words—it was finally time to open the “vault.”

***

Thanks to the user "" for the chapters provided!