# AI Labs Will Trap You or Replace You - Dillon Rolnick, Nous Research

Dillon Rolnick, Nous Research

[Talk page](https://ai.engineer/talks/Sg6vtKOAOQ0-ai-labs-will-trap-you-replace-you) · [Video](https://www.youtube.com/watch?v=Sg6vtKOAOQ0)

## When Models Get Cheaper, Who Owns the Investment Process?

Dillon Rolnick, CEO of Nous Research, argues that falling model margins give labs a reason to raise prices on firms that depend on their agents or compete with those firms directly. An investment decision illustrates what a firm can keep under its control when it builds on an open framework.

### Falling token prices complicate a trillion-dollar bet

[0:12](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=12s)

Dillon Rolnick opens with a question about recovering the cost of building AI. He puts individual labs’ infrastructure and research bets in the hundreds of billions of dollars and the industry’s combined investment at perhaps trillions. Consider the customer on the other side of that bet. An engineer curious about a new model can move among Astra, Fable, Opus and Kimi with little attachment to the company serving the request. Each lab has paid to develop a capability that its customer feels increasingly free to shop around for.

Rolnick points to three trends in charts that had circulated that week. The average price of a token was falling, total spending on tokens was surging, and the infrastructure costs associated with serving them had stayed roughly flat. Growing demand and shrinking margins can coexist. Customers can buy far more tokens even as competition narrows the premium a lab can charge for a particularly capable model. In Rolnick’s reading, a broadly available baseline of intelligence changes where a lab can expect to make its money.

### The agent layer gives a lab leverage over business output

[1:53](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=113s)

Rolnick sees two broad responses. A lab can move to another layer of the stack, either down toward infrastructure or up toward the user. Or it can win the race to what he calls godlike intelligence and monetize the lead. He spends most of his time on the routes that bring a lab into a customer’s business. Owning the interface puts the lab in a position to retain the user. Developing a sufficiently capable system gives the lab a reason to pursue the user’s business itself.

![A two-item numbered list appears in white over a blurred blue image of a figure seated cross-legged.](https://ai.engineer/api/talk-essay-media/a63510dcd5d7f47850f61de89712833ab0076e542288752eac10c61b11a0adb4)

Compete at a different layer / Build god

Agents make the first route especially attractive. Rolnick considers consumers fickle, while businesses buy large volumes of tokens for work that generates economic value. He recalls an earlier speaker mentioning 5 percent of a workforce engaging with agents. He asks executives to consider a different measure two or three years out. What share of the company’s output might be agentic? His own forecast is 70 percent, which he considers conservative. The two percentages describe different things. A head count of agent users tells an executive little about how much production depends on their tools.

Rolnick likens the resulting dependency to someone else having the power to organize a strike. If an outside platform coordinates most of a firm’s output, its owner can change the terms once the firm has embedded it. A rival’s cheaper tokens offer little immediate relief to a company whose work runs through the incumbent. The possibility of interruption gives the provider room to raise prices.

### A lab that can solve a market has a reason to enter it

[3:55](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=235s)

Take the second route and suppose a lab achieves the intelligence it set out to build. It now has a choice about a lucrative problem. It can sell intelligence at a narrowing margin to a customer who captures the return, or it can direct that intelligence at the problem and collect the return itself. Rolnick believes labs view financial markets as tractable enough to make the latter course tempting.

Domain knowledge constrains that ambition today. A lab may believe it can solve markets and still lack the context, know-how and experience that a financial firm brings to an investment decision. Rolnick points to labs hiring investment banking analysts and partnering with some of the world’s largest firms. Those moves give labs access to expertise they lack. His prediction is that they will eventually pursue the financial opportunity directly. For a firm considering a partnership today, that possibility matters before a competing product exists.

### The cost of leaving grows as the organization learns

[5:18](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=318s)

Embedding a product and raising prices once customers depend on it is a familiar strategy. What changes, Rolnick argues, is the scale. He believes tens of percent of the world’s economy could be at stake. For the executives and engineers choosing a platform today, dependence grows as their colleagues learn how to get good results, turn that knowledge into routines, and build systems around those routines. A later move has to account for the know-how the firm has built into those systems.

That is why Rolnick advocates choosing an open foundation while the processes are still taking shape. He recommends Hermes Agent, then addresses his own company’s incentives. Nous Research, he says, began as an effort to resist concentrated control of a technology its founders considered exceptionally consequential. He describes a staff of science-fiction fans for whom the risk was easy to picture. His joke that Skynet was closed source makes the founders’ preference clear. The business question for Nous is whether it can earn a living serving customers who retain the power to leave.

### Nous reproduced a Grok bots interface in a weekend

[7:57](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=477s)

When xAI released Grok bots, the Nous team liked how intuitive the interface was. Rolnick says they pointed Hermes Agent at it, reproduced the interface over a weekend and released the result as an open-source feature of Hermes Agent. The concrete comparison is the time it took another team to deliver a user experience it admired. Rolnick gives no performance comparison between the underlying systems.

He contrasts the weekend’s work with the economics of building SaaS. Assembling open-source libraries into a polished proprietary product used to require enough coordinated engineering labor to protect the result for a time. Agents also reduce the labor required by motivated open-source developers working in the other direction. A closed supplier can still introduce a desirable feature. Rolnick expects the interval during which it alone can offer that feature to shrink. He sees more durable advantages in network effects, reliability and reputation.

### Distillation puts pressure on a model’s lead

[9:13](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=553s)

Rolnick sees a parallel in model development. In distillation, a capable model supplies outputs that serve as supplementary training data for another model. In his account, the student can approach the teacher’s capability because much of the teacher’s advantage comes from data quality and training scale. A competitor with sufficient scale can learn from the teacher’s examples even without reproducing a proprietary architectural invention.

He describes two responses from labs. They characterize unauthorized distillation as theft, and they train models to decline assistance with training other models, a restriction they describe as safety. Rolnick challenges both framings. In particular, he sees a tension between claiming ownership over learning from a model’s outputs and having trained that model on people’s written and creative work. His argument concerns the incentives behind those positions. He offers neither a legal ruling nor a benchmark of student performance.

### An investment decision shows what the agent coordinates

[11:06](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=666s)

Rolnick recalls an earlier presentation about the inputs a financial firm brings to an investment decision. They include data sources, routine code, people’s opinions and judgment. An agent framework lets an application combine deterministic operations with reasoning whose outcome can vary. Routine code can process an input in a repeatable way. The application can then bring that result together with an analyst’s view when the decision calls for judgment.

This is what Rolnick means by comparing an agent to an operating system or an application framework. A personal assistant exposes the interaction a user has with it. The framework also determines how code, evidence and reasoning fit together underneath. His architectural test is whether the firm can swap the model while retaining control of that arrangement. A new model may reach a different judgment. The firm should still decide which sources and opinions it receives and where that judgment enters the investment process.

![Five phrases stand out in white boxes against a blue paragraph. They highlight using local models, switching models mid-conversation, running offline, reading the prompt, and exporting the agent.](https://ai.engineer/api/talk-essay-media/ac1b53615da70f612eae294c0e228b741e79e912eb6f9b01565add0c482280db)

What you should be able to do with your agent

### Keep the investment process and inherit the framework improvements

[12:50](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=770s)

Ownership creates a maintenance question. A financial firm has reasons to develop its investment process that do little to justify spending its engineering effort keeping a general-purpose agent framework current. Rolnick’s proposed division of labor lets the firm build increasingly sophisticated applications from code, data and reasoning while inheriting improvements to the common machinery. The firm can concentrate on the expertise that made it successful as the state of the art in agents changes.

Hermes Agent is Nous’s offer on those terms. Rolnick says customers can choose where and how to run it, modify it, and keep using it if they end their relationship with Nous. He wants it to become a standard and describes it as the most used and downloaded open-source agent at the time, with far more downloads than some closed competitors. He gives no counts or measurement method. The commercial proposition is more concrete. Customers who hold the software themselves can still pay Nous for support, safety and help operating at scale. Nous then has to earn their continued business through the quality of that work.

![A network of glowing points branches above a small figure. The Hermes statement is centered, with GitHub stars and repository rank at lower left and user and clone figures at lower right.](https://ai.engineer/api/talk-essay-media/fa6f1d75d5090f1328971106bf5a714a88d4af617205a1451daf05665d877365)

Hermes is the open standard for agents.

## Key takeaways

- Measure the share of business output that depends on an agent platform. Employee adoption and token spend alone can obscure what a change in the provider’s terms would put at risk.
- Trace one investment decision through its data sources, routine code, analysts’ opinions and model judgment. Check which parts the firm could carry forward if it changed models or left the provider.
- Separate the work that encodes the firm’s expertise from general agent maintenance. A framework can supply improvements while the firm controls its process, deployment and choice of model.

## Slides

### The model is commoditized

[0:55](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=55s) [Centered white text overlays a grainy blue image of a row of similar figures, with one figure raised above the others.](https://assets.ai.engineer/sites/aiecode2025/sha256/5e3db090f27ea439cacbe8b562e960a44268f54a7711f0986db8bc590f9687d3.webp)

// NOUS RESEARCH

The model is commoditized

NOUS

### Compete at a different layer / Build god

[1:53](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=113s) [A two-item numbered list appears in white over a blurred blue image of a figure seated cross-legged.](https://assets.ai.engineer/sites/aiecode2025/sha256/a63510dcd5d7f47850f61de89712833ab0076e542288752eac10c61b11a0adb4.webp)

// NOUS RESEARCH

1.Compete at a different layer
2.Build god

NOUS

### Point the models at the most lucrative problems in the world

[3:59](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=239s) [Four centered lines of white text describe a strategy for earning money back. A blurred dark shape spans the blue background.](https://assets.ai.engineer/sites/aiecode2025/sha256/9d611dcfb0e493ea3adee9c739713fc113ec14c07a67279de6484b9e60d99d71.webp)

// NOUS RESEARCH

The best shot at earning it
back is to point the models
at the most lucrative
problems in the world.

NOUS

### Labs think markets are solvable problems

[4:26](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=266s) [A two-line statement about how labs view markets is centered over a blurred blue skyline.](https://assets.ai.engineer/sites/aiecode2025/sha256/3ddabfac5f70e4257afa69aa112eaa1b9d7399c36df42099983f0b87f3309c34.webp)

// NOUS RESEARCH

Labs think markets
are solvable problems.

NOUS

### Having alpha over the rest of society in seeing the future

[4:40](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=280s) [A white speech bubble points toward a blue-tinted portrait of a man in a sweater. The quotation describes groups of people who can foresee the future better than others.](https://assets.ai.engineer/sites/aiecode2025/sha256/658e46e7f9859ebb32f15a3c73d246695f7ceadce78dcee5e7e5a574af940775.webp)

// NOUS RESEARCH

There can be... groups of
smart people... who do just
have alpha over the rest of
society in seeing the
future.

NOUS

### Big labs underestimate their own tools and users

[7:53](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=473s) [Three lines of white text appear below a silhouetted figure standing beside a small glowing screen in a grainy blue scene.](https://assets.ai.engineer/sites/aiecode2025/sha256/189b6c17db6e66bf79943b4e20ea39e61490d173bd319e9d7229de51950734de.webp)

// NOUS RESEARCH

Big labs underestimate
their own tools and
users

NOUS

### “Agents” are operating systems

[9:54](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=594s) [White text on the left overlays a grainy blue image of two hands reaching toward a square object in the center.](https://assets.ai.engineer/sites/aiecode2025/sha256/73d691e499459cc5838c2e63fb3359122359f810f34d8bc1c21376d8e07d64a1.webp)

// NOUS RESEARCH

“Agents” are
operating systems

NOUS

### Water flows downhill to open source.

[9:59](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=599s) [A pale stream winds from the upper right to the lower left through dark blue terrain. The slide text sits beside it at lower right.](https://assets.ai.engineer/sites/aiecode2025/sha256/25a51b018a9d6f46e11a71d33cc36e852cdf1be7ef3ae4306475101c9f287d3d.webp)

// NOUS RESEARCH

Water flows
downhill to
open source.

NOUS

### What you should be able to do with your agent

[12:21](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=741s) [Five phrases stand out in white boxes against a blue paragraph. They highlight using local models, switching models mid-conversation, running offline, reading the prompt, and exporting the agent.](https://assets.ai.engineer/sites/aiecode2025/sha256/ac1b53615da70f612eae294c0e228b741e79e912eb6f9b01565add0c482280db.webp)

You should be able to use smarter models in your agent. You should be
able to use cheaper models in your agent. You should be able to use
local models in your agent. You should be able to use a different
model for every job. You should be able to switch models in the
middle of a conversation. You should be able to choose what data
your agent has. You should be able to choose what it remembers and
what it forgets. You should be able to choose what compute your agent
runs on. You should be able to run it on a machine that never
touches the internet. You should be able to reach it from the apps
you already use. You should be able to choose how your agent thinks.
You should be able to read the prompt it runs on. You should be able
to see every tool it has, and turn off the ones you don't want. You
should be able to teach it something once and never explain it again.
You should be able to know what it did, and when, and why. You should
be able to export your agent.

### Hermes is the open standard for agents.

[13:46](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=826s) [A network of glowing points branches above a small figure. The Hermes statement is centered, with GitHub stars and repository rank at lower left and user and clone figures at lower right.](https://assets.ai.engineer/sites/aiecode2025/sha256/fa6f1d75d5090f1328971106bf5a714a88d4af617205a1451daf05665d877365.webp)

// NOUS RESEARCH

Hermes is the open
standard for agents.

250K GitHub stars
15th most starred repository ever.

Millions of users.
70M+ lifetime clones.

NOUS

## Full transcript

[0:12](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=12s) I'm Dylan Rolnik. I'm the CEO of Nous Research, and today we're gonna talk about building on foundations and platforms that you can actually own.

[0:29](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=29s) Oh, there we go. All right. How do you make back a cent-a-billion-dollar infrastructure and research bet? Labs individually have spent hundreds of billions of dollars training and building ever more sophisticated intelligence system, and the industry as a whole has spent trillions perhaps.

[0:56](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=56s) The problem is, though, that the models have begun to commoditize. You guys in this room who are hopefully practitioners of AI have known this intuitively. You switch among models readily. You switch between Astra, Fable, Opus, Kimi, whatever it may be, without too much regard for who's on the back end or what it may be. You wanna try the new model. We've seen this in data too. There was some charts going around Twitter this week showing

[1:26](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=86s) the average cost of a token going down while the total spent on tokens surges, and the infrastructure costs associated with serving those tokens has stayed flat. Meaning, where there used to be margin in a very, very performant model, it's now starting to narrow as we see competition and, you know, a baseline level of intelligence be available to people.

[1:53](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=113s) Okay, so if the models are starting to commoditize and you've spent hundreds of billions, trillions of dollars on them, you need to make that back, and you need to find a different way to do it. And in my opinion, there are two pretty obvious ways to do it. One, you find a different layer of the stack to compete at. You go down to the infrastructure, or you go up to the user. Or two, you build God, and you build God faster

[2:23](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=143s) than anyone else, and you monetize God.

[2:27](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=147s) If you're in front of the user, you have an opportunity for stickiness. Stickiness means margin. This is why the fight has largely shifted to the agent layer lately. We started talking-- I heard somebody previously talk about five percent of, you know, a workforce engaging in agents. This is becoming more and more the zeitgeist of where the AI battle is taking place. Consumers, though, are fickle, and they're terrible customers. Businesses, on the other hand, spend tons of money

[2:57](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=177s) on tokens, and they have economically productive use cases for those tokens. So if you're going to be competing at the user layer and you're going to be bringing this fight somewhere else, you're going to be bringing it to businesses. The best thing to do is to get ingrained in a business 'cause when you're in a business, it's hard to take you out, right? And if you think about your business two, three years into the future, how much of your workforce, as measured by its output,

[3:27](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=207s) is agentic? Seventy percent conservatively, probably more. And if seventy percent of your workforce is organized by somebody else who may or may not have adverse incentives to you, they can be organized into a strike. And so what you would do if you had to make back a trillion-dollar investment was you would embed, you would use your asymmetric leverage, and you would change the rules of the game.

[3:55](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=235s) Or, like I said, you would build God.

[4:00](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=240s) If you succeed at building God, what you would do is you would point God at the most lucrative problems in the world.

[4:10](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=250s) The best shot at earning it back is not to sell it for an ever-decreasing margin to others to solve these hard and lucrative problems. The best shot is to solve them yourselves.

[4:28](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=268s) Labs think that markets are solvable problems, and that's exactly how they view the financial industry, as a problem to be solved,

[4:41](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=281s) and a tractable problem at that, perhaps not even a very hard one. Right now, though, they know, at least the smart ones know, that they don't have enough context, know-how, or experience in the financial industry to be able to do that themselves, so they will partner. They hire investment banking analysts. They partner with some of the biggest firms on the globe, and they start to build their foundation here. My belief is that they will eventually go to

[5:11](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=311s) the source because, like I said, that's where the money is, and we're trying to make back a trillion dollars.

[5:18](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=318s) And so you might say this is nothing new. We have always been trying to trap people into these games and change the margin, and once we have you and you're stuck, we are able to raise the prices. But the difference is that the stakes are much higher. We're talking about meaningful deca percentages of the Earth's economy. And you in this room

[5:42](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=342s) need AI. Everybody has seen what they can do, what these models can do, what agentic systems can do, and there's no going back.

[5:51](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=351s) People in this room who are, um, making decisions, vice presidents and executives and engineers, about what they're building on are going to have to make decisions now about where they grow roots around these tools and processes that will affect the future of these businesses.

[6:11](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=371s) And whatever you pick, you'll start to harden. You'll gain know-how. It'll sop up the intelligence of the organiz- organization itself. You will start to build processes and systems around it. I implore you, do not build on quicksand found- quicksand foundations. Build on open foundations. Build on Hermes Agent.

[6:33](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=393s) So I just spent a fair amount of time convincing you, hopefully, that the highest expected value move in AI right now is to either trap you and jack up the prices or replace you directly. So why am I standing here? Uh, we view the world very differently. And the first thing I hope you guys will learn about me and about Nous Research is that we built this company primarily for moral and philosophical

[7:03](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=423s) reasons. And a lot of times in business, morality and incentives don't always align, and when that happens, it's probably easiest to change your morality. But we think the world is changing, and we think it's different. And for that reason, we think that the best course of action is one in which we are both aligned.

[7:25](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=445s) When we built the company originally, we built it to combat a seemingly inexorable trend towards centralization and centralized control of perhaps the world's most inte- most important technology and most important invention. Nous Research was built by and staffed by nerds, nerds who love science fiction. And we've all seen what happens when a centralized entity gets control of godlike intelligence and uses that for exploitation. Skynet was, after all, a closed source company.

[7:57](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=477s) The next thing that you need to know is that there's no longer really a strong moat in proprietary software. The marginal cost to reproduce a piece of software has plummeted. A few months ago, uh, xAI redu- released Grok bots, and we saw that and we thought, "Wow, that's an intuitive user interface. That's really nice." So we pointed Hermes Agent at it, and we rep- reproduced it in a weekend and released it as a feature within Hermes Agent, open source, of course.

[8:27](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=507s) And the game has changed. It used to be that software was difficult to produce, and it took coordinated teams of engineers long hours to pilfer open source and take libraries and combine them together into proprietary pieces of software and then sell that software as SaaS. But now the opposite is true. Anything can be reverse engineered, and there are many, many motivated open source builders pointing their agents at these different softwares. If the

[8:57](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=537s) closed competitors do manage to make interesting features, open source will have it not long afterwards. The moats of yesterday were in proprietary IP. The moats of tomorrow are in network effects, reputation, and reliability.

[9:13](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=553s) What's funny is that in a lot of ways, the labs have always seemed to underestimate their own tools, or perhaps they are underestimating the users of those tools and their genius and their motivation. This happened and is happening on the model side, and we'll talk about why that's important. But it turns out that models can fairly easily teach another model to be, you know, nearly as smart as the original model. And the advantage in a model is

[9:42](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=582s) very rarely in the architecture or some novel system that was created in a proprietary way for that model. In fact, it's usually in the quality of the data and the scale.

[9:54](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=594s) And  sorry.

[10:05](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=605s) The, uh,

[10:13](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=613s) The, the model started to re- the model companies started to realize that the, you know, scale was catching up and that you could create these things with supplementary quality data by teaching one model from another. So they labeled the teaching process distillation, and they labeled illicit distillation theft. You cannot call that copyright infringement because copyright infringement would imply that you somehow took

[10:43](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=643s) every living human being's written or creative work and then trained on top of it. They also labeled the process of-- They also trained the models to not assist in the process of training other models, and they labeled that safety.

[11:06](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=666s) The next thing that you need to know about agents and about the new world is that agents are like operating systems, and they're like frameworks, and they're a lot more like those than they are like chatbots. And so if you continue to think about your agent as merely a personal assistant, you will forever live on the surface of what it means to engage in agentic systems. And so what do I mean by that?

[11:31](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=691s) If you guys were here a little bit earlier, there was a presentation about a financial firm and how they think about AI and these different inputs that go into ultimately what becomes an investment decision. Agents are really at their heart a system that combines determinism and non-determinism. If you remember from that talk, there are data sources, there are routine code, there are people's opinions, and there's judgment. And all these things come together to create what is the new application. The new

[12:01](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=721s) application is no longer necessary that it is deterministic and solid and cold, hard code. It can inject intelligence, reasoning, judgment, all of these things together to combine to create The new application. And so as you at your company

[12:22](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=742s) think about building these new applications, you should be building these systems on foundations that you own. You should be able to swap in and out the models that you want. You should be creating ever more sophisticated towers of functionality on top of these core primitives of determinism, non-determinism, code, data sources, judgment, reasoning, intellect.

[12:50](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=770s) But you should not be spending your entire lives being intelligence companies. What has made you successful this far? Most of you are most likely not agent companies. Most of you are not likely AI companies. You should be focused on the things that you do well, and when the state-of-the-art changes and it evolves and what it means to be an agentic system starts to shift, you should inherit that, not rebuild it and maintain it. Remember, these are tools that should make you better. They are not

[13:20](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=800s) drop-ins to replace you. Do you know how much energy it takes to create a human?

[13:27](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=807s) And the beauty of the Hermes agent proposition is that you can run it however you want, wherever you want, modify it as you see fit. If you're not having fun anymore, you can take your ball and you can go home. The moat has changed. The moat is no longer entrapping you. The moat is in reputation, and that's what Hermes agent offers you. And so what do I want out of all of this? Well, I want to build a standard. Hermes agent today

[13:57](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=837s) is the most used and downloaded open source agent in the world. It is order of its, of magnitude more downloaded than some of its closed competitors. And I want everybody to be able to build on these open foundations because as long as one of us, us, is willing to open source these tools and these agentic systems, water will continue to flow downhill. And I want to be downhill so that I can work with you to provide support, safety, scalability to go along with the world's best open

[14:27](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=867s) agent. I've seen where this goes, and we've seen where the moats will lie, network effects, reliability, and reputation.

[14:37](https://www.youtube.com/watch?v=Sg6vtKOAOQ0&t=877s) People are going to continue to solve hard problems with agentic systems. People are going to solve lucrative problems with agentic systems. Somebody is going to earn back the trillion-dollar investments that we've made. Make sure that it's you. Make sure that you are in control of your own destiny. You can have your cake and eat it too. You can have state-of-the-art features and reliability without sacrificing your autonomy. Build on something you own. Build with us. Thank you.
