Exploring the Advancements and Challenges in AI Agent Deployment
John Nay, founder and Chief Executive Officer of Norm Ai, joins theCUBE's special presentation with NYSE Wired, focusing on the upcoming Artificial Intelligence Agent Conference 2025. Hosted by John Furrier, co-founder and co-Chief Executive Officer of SiliconANGLE Media, this insightful discussion covers the pivotal developments in AI infrastructure and the regulatory complexities faced by enterprises.
In this episode, Nay shares their expertise in regulatory AI infrastructure, particularly as it pertains to AI agent deployment in highly regulated sectors. The conversation, hosted by Furrier, delves into the evolving landscape of AI technology, compliance challenges, and the strategic initiatives underway at Norm Ai to address the pressing issues surrounding AI deployment. The discussion provides valuable insights for both technology and policy influencers.
Key takeaways from the discussion include the emphasis on the need for dynamic, real-time compliance frameworks that align with regulatory standards, as emphasized by Nay. Furthermore, the episode highlights how enterprises can leverage existing compliance structures to integrate AI technologies more effectively, offering a glimpse into the future of AI agent scalability and regulation. The conversation underscores the importance of bridging the gap between engineering, policy, and technology for sustainable AI innovation.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: AI Agent Conference. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: AI Agent Conference.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: AI Agent Conference
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: AI Agent Conference.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: AI Agent Conference. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: AI Agent Conference.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: AI Agent Conference
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: AI Agent Conference. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
AGNT Podcast Ep. 6 with Gemma Allen & Raphaëlle d'Ornano
Exploring the Advancements and Challenges in AI Agent Deployment
John Nay, founder and Chief Executive Officer of Norm Ai, joins theCUBE's special presentation with NYSE Wired, focusing on the upcoming Artificial Intelligence Agent Conference 2025. Hosted by John Furrier, co-founder and co-Chief Executive Officer of SiliconANGLE Media, this insightful discussion covers the pivotal developments in AI infrastructure and the regulatory complexities faced by enterprises.
In this episode, Nay shares their expertise in regulatory AI infrastructure, particularly as it pertains to AI agent deployment in highly regulated sectors. The conversation, hosted by Furrier, delves into the evolving landscape of AI technology, compliance challenges, and the strategic initiatives underway at Norm Ai to address the pressing issues surrounding AI deployment. The discussion provides valuable insights for both technology and policy influencers.
Key takeaways from the discussion include the emphasis on the need for dynamic, real-time compliance frameworks that align with regulatory standards, as emphasized by Nay. Furthermore, the episode highlights how enterprises can leverage existing compliance structures to integrate AI technologies more effectively, offering a glimpse into the future of AI agent scalability and regulation. The conversation underscores the importance of bridging the gap between engineering, policy, and technology for sustainable AI innovation.
AGNT Podcast Ep. 6 with Gemma Allen & Raphaëlle d'Ornano
Gemma Allen
Host, theCUBE + NYSE WiredtheCUBE
HOST
Raphaelle d'Ornano
Founder & CEODecoding Discontinuity
In this interview from the AGNT Podcast, recorded at the MCP Dev Summit in Times Square, Raphaelle d'Ornano joins theCUBE + NYSE Wired's Gemma Allen to break down the week's most market-moving developments in AI — from the accidental leak of Anthropic's 512,000-line Claude Code repository to the misread implications of Google's TurboQuant research. D'Ornano unpacks what the leaked codebase actually reveals: a sophisticated orchestration graph with Claude Code as the entry point for capturing user intent, and an indexing mechanism called the Pointer that solve...Read more
exploreKeep Exploring
What are the implications of the Claude Code leak at Anthropic — what happened, was it a joke or an actual leak, and what does it reveal about their technology (e.g., orchestration graph and the Pointer system for managing context)?add
What significant innovations and capabilities did Anthropic reveal in its leaked/demo materials?add
What would someone in Sam Altman's position likely have thought when the leak of a massive (≈512,000-line) GitHub repository and reports of agents being told to pretend not to be agents became public — especially regarding governance, product integrity, enterprise reputation, and capital‑markets implications?add
Will the recent leak/security incident delay their IPO plans or cause reputational damage?add
Why did Google's TurboQuant research trigger a sell-off in memory stocks, and does TurboQuant actually reduce long‑term memory demand for AI infrastructure?add
Can OpenAI realistically pursue both a consumer/ad-driven model and an enterprise-focused, vertically integrated model at the same time, or should it focus on one?add
AGNT Podcast Ep. 6 with Gemma Allen & Raphaëlle d'Ornano
search
Gemma Allen
>> Welcome to AGNT, the podcast where enterprise tech meets the authentic era. I'm Gemma Allen, joined by my co-host, Raphaelle d'Ornano, broadcasting from the New York Stock Exchange. In every episode, we unpack how intelligent systems are reshaping companies, markets and the way real work gets done. From Fortune 500 boardrooms to breakout upstarts, we're digging into the strategies, technologies and people defining the next chapter of AI. Let's get into it. Welcome to a very special episode of AGNT. We are not on Wall Street today. We're in Times Square in New York at the MCP Dev Summit, a special guest of the Linux Foundation. Raphaelle, welcome.
Raphaelle d'Ornano
>> Thank you.
Gemma Allen
>> I think when I look back on my own analysis of MCP, yours were some of the first articles that I ever read. So this is really somewhat of a full circle moment for you, right?
Raphaelle d'Ornano
>> Right.
Gemma Allen
>> Lots to cover today, but let's get straight into it. News of the week, the Claude Code fiasco. I don't even know how to term it yet, because I think the implications are yet to be fully known. What are your thoughts?
Raphaelle d'Ornano
>> Well, look, I think... So first, we all saw this was April Fool's joke, right? It came out on April 1st, though it officially was done on March 31st, but everyone had a hesitation like, "Is this a joke?" You have Anthropic, this major company that is going to IPO and that has said just last week or this week that it's going to IPO this year. And then you have this, I don't even know the term, huge monster, whatever leak where it's a completely random leak. Random, I don't know if that's the term, but of course it was not supposed to be done. And they unveiled, I think it was 512,000 lines of code-
Gemma Allen
>> Wow.
Raphaelle d'Ornano
>> Where basically you have a very strong view or preview, whatever is the right term, of what they are building. And what Anthropic is building is what makes the stock market shake each time. Over the past two months, every time Anthropic releases something, the stocks drop. Anthropic releases Cowork, it drops, the plugins, the Claude Code security. So now the question everyone is asking is like, "Okay, what was in there that matters?" I think several things. First, Anthropic is really confirming that they are building what I have termed this orchestration graph.
Gemma Allen
>> For sure.
Raphaelle d'Ornano
>> They want to be the orchestrators with Claude Code as where you capture user intent, the whole orchestration system that becomes a very sophisticated agentic construction. And if there's one thing that I want to point out, it's how they are solving the problem of context. Context entropy, which is the name of the problem, which is as these models absorb more and more context to actually be able to render the objectives, at one point the model performance goes down and you're not able to handle context even though the context windows have expanded, expanded. So they have apparently built in a very smart system called the Pointer by which, instead of actually having to put the information within the loop, you can go and index the information from everywhere it is, so from the database, from the CRM from... Think of this as a very smart knowledge management system. This is what the LLM becomes. You have the model and you can go and fetch the information where you need it, when you need it, without it taking a lot of space in the context window. That seems very technical. That is a major innovation. And if Anthropic is effectively able to do that, they completely bypass all the competition. All the competition.
Gemma Allen
>> I think it's really interesting that, even though, yes, this was a huge screw-up on behalf of someone at Anthropic and I would hate to be part or parcel of that team that was responsible for that. Can you imagine? When people were thinking it's an April Fool's joke, I was like, "Dario doesn't strike me like somebody with that type of sense of humor." Why, right?
Raphaelle d'Ornano
>> Why?
Gemma Allen
>> But at the same time, so much of the conversations are around catch up, catch up, get ahead. It just goes to show how much impetus they have in this market. What I thought was really interesting from what we saw, like we saw Capybara, we saw Tamagotchi-style friends, which you can engage with, it was covering so many elements of not just the enterprise stack, but also the consumer stack too. But I thought it was really interesting that a lot of it was about persistent activity in the background all the time, which is in some ways similar to the OpenClaw, now NemoClaw argument, right?
Raphaelle d'Ornano
>> Completely.
Gemma Allen
>> The idea that Anthropic can do this for you too, perhaps in a way that just, at least from a perception perspective, has more guardrails, more governance. And that's a really interesting part of this because we all wondered about what happened with that Peter, Dario conversation some months back. So perhaps was this in their plan all along? It's just so fascinating.
Raphaelle d'Ornano
>> For sure that person is in a lot of trouble and, of course, all of the open source community went big into this. They had actually put in the features, everything against distillation. So they had found a way to put poison data if you are distilling the model. So if you have open source Chinese competition that wants to distill you, Anthropic has been saying that people are distilling their models, they actually put a system that was unveiled by which they would poison the data of people who would distill their system. So it's unveiling that. So it's a problem. I think to go back to your point on operating in the background, that is indeed, I would say the second big innovation that they unveiled, that they unveiled unexpectedly, which is having this agent system work in the background. You leave on Friday and you arrive on Monday morning and everything has been done for you. You have the notifications. They're really working to have these agentic colleagues, as we have used often this term, to be working for you. And I think that what they show, if you put everything together, it's not just small details. It's a very complex architectural construction by which they're going all into this agent harness, multi-agent orchestration. What we have been saying since May last year, when we said the moat of LLMs is not in technical capabilities only, it's in the orchestration. This was in May 2025. Anthropic has been going along that line every single day and building orchestration, orchestration, orchestration, and now it's really going into multi-agent systems, running all the time. Claude Code is really the entry point. And so it has interesting implications that we can discuss in another episode on every company.
Gemma Allen
>> Let's quickly talk implications though, because a lot of headlines yesterday picked up this idea that from a governance perspective and from a integrity perspective, it raised some questions. There was some commentary around agents being told to act like they weren't agents to access certain systems, et cetera. And I feel like perhaps that will give them somewhat of a grit on the enterprise veneer that they've built. But overall, if you were Sam Altman yesterday, what do you think it means? Where do you think his mind went yesterday when this news broke?
Raphaelle d'Ornano
>> What I think is that it just shows you the pace at which this is happening. Imagine when they leaked, if that's the term, you have people in Asia that, at four o'clock in the morning, were building this huge GitHub repo that is the fastest growing GitHub repo in history by the way.
Gemma Allen
>> Incredible.
Raphaelle d'Ornano
>> People are all watching what is happening, but it's happening... We're not just talking about they leaked one feature or two features. No, it's 512,000 lines of code with the biggest player going to IPO. I always think about the implications in terms of capital markets. This is not just a random mom and pop shop company. This is the company that everyone is watching.
Gemma Allen
>> Do you expect it will in any way delay their IPO plan? Do you think they're going to have a reputational overturn from this?
Raphaelle d'Ornano
>> No. Look, I may be wrong on that. I think that there was no customer data that was leaked. The model weights and all that were not leaked. So again, it raises questions about, "Oh my God, how can that happen?" The funny part is it actually happened through an acquisition that they did in December 2025, if I'm correct. So it's actually through something that they had bought. It's with that system that the leak happened. Again, I don't know all the technical details, but it shows. So I don't think that that is going to affect the power of what they're building and the direction. For me, it's the direction that matters. And I think the Tamagotchi part was maybe more anecdotal, but I'm still focusing on what are they building in the enterprise and why does this matter, and thinking what does this mean for Snowflake, Databricks? What does this mean for Salesforce, Workday, ServiceNow? What does this mean for Palantir, all these companies?
Gemma Allen
>> They've had an interesting run from a geopolitical perspective too, and they have had a large influx of consumer customers. Subscriptions have soared for Anthropic on the back of what's happened here in the U.S.
Raphaelle d'Ornano
>> Exactly.
Gemma Allen
>> So I think it's an interesting time to see exactly what direction they're going, but one thing is for sure, I think Dario is a genius and I think we can't deny how successful and how impactful this company is. The fact that this leak has... People have probably not gone to bed. Can you imagine-
Raphaelle d'Ornano
>> Oh, for sure. For sure.
Gemma Allen
>> In the dev community what's happened since? So I think it's fascinating, but let's move on for a second because this was a crazy week. We talk all the time about market reaction to tech and how we live in a moment that feels like a bit of a pied piper moment. No one really knows what's happening. I think TurboQuant is another exceptional example of this. So this is the memory issue that happened this week whereby Google released this report that basically demonstrated how caching can essentially be compressed. And I think Wall Street heard memory and memory stocks just got an absolute wipe-out. It is insane. If you think about it, I heard someone describe it. It's like Google came up with a way to fit more content onto a Post-it note and filing cabinets lost 30% of their market value.
Raphaelle d'Ornano
>> Oh, my god.
Gemma Allen
>> So it's a crazy, crazy, crazy space. What are your thoughts in terms of what's happened and why these two worlds are not connecting with sensical data?
Raphaelle d'Ornano
>> Well, so I think we need to take a step back on memory as a whole. And the Mag7 had a run over the past years as foundational players, of course, in the tech ecosystem. I think memory is maybe the equivalent of the Mag7 trade for a lot of hedge funds right now. Why? Because memory in AI infrastructure is a real bottleneck. From a supply side, there is no more capability until 2027, maybe 2028. The delays vary, of course, according to the producers. You have three companies in the world, two in South Korea, Micron in the U.S. So SK Hynix, Samsung and Micron in the U.S. that are basically having an oligopoly on HBM, for example. Sandisk is producing NAND so those are the key players. And a lot of hedge funds are invested in these stocks because there is this bottleneck. And Wall Street loves bottlenecks because a bottleneck, you can increase prices, increase prices, and of course, that's a favorable investment thesis. So when you have a technique that comes out where it's like, "Oh, we're going to need six times less memory," well, all the quants are like, "Memory equals less demand equals problems." So of course this is not what happened because the constraint is from a supply side. And what happened from Google, so two things. First, the paper was released under a research blog and points back to three research papers that are from 2025 and one from 2024. So they packaged it in the right way because they're presenting this at an upcoming AI conference in the next week, but the actual papers on TurboQuant are actually from one year ago. That's the first part, so it's not that new. It's been out there for a year. Second, this concerns only inference efficiency. And we go back here to the long context that we were talking about with the Claude leak. Long context is key and being able to solve the memory constraints around how long context is able to be done cheaply, quickly, efficiently is a key problem that Google is solving. So everything was mixed up between the memory for the model waves, the memory for the KV cache indeed. And I just think that this is a very interesting architectural innovation that is going to actually accelerate the need for memory and that... Anyways, everything is accelerating this inference economy that we have mentioned many times. So I think it's a net positive. And of course, you cannot put all of the eggs in the same bag and just confound everything.
Gemma Allen
>> Do you think that at a hedge fund level... I'm interested to know your perspective on this, I'm sure there's people who are extremely informed, but broadly as these narratives play out and they have market impacts, do you think people understand model waves, KV cache, neural mesh networks, all of that stuff? Do you think that the... Because memory is a very large space and it is a very important space. It has a huge growth. If you think about what it does now in the inference era of AI-
Raphaelle d'Ornano
>> Exactly.
Gemma Allen
>> Versus the training model era of AI, versus what it did like 20 years ago, people just imagined hard drives. The technicalities of this are huge. Do you think that people are translating those technicalities into facts? Where is the mismatch of understanding happening? That's what I would really love your perspective on.
Raphaelle d'Ornano
>> Well, to some extent, that's what I'm doing and this is what I invest behind. My conviction is that architectural resilience. And so when I say architectural resilience, it's what is happening in the AI stack and how are you positioned as a memory company, as a GPU company, as an application company, how are you positioned versus an innovation? And what does that mean? What is really complex is this is super, super technical, yes, and this is happening so fast. I think, again, just go back to the Claude leak, this happened two days ago. Here we are talking about that. There's billions at stake everywhere. We're just starting to process that and it happens every single day, several times per day. So I think, in this new paradigm of AI, agentic AI that I think no one is questioning anymore. There are problems with how it is built out, the power constraints, the cost of this huge build-out. So it's full of question marks for sure, but I think the progress, what is being done, what Anthropic is doing, no one is questioning that anymore, hopefully. And so, the investment paradigm is changing. And we're living through that right now. That has been my fundamental conviction over the past two years and it's complex, but it's fascinating.
Gemma Allen
>> And lastly, because I know I really want to move on to OpenAI, who is responsible to do a better job, do you think, at messaging these product roadmaps and impacts? Do you think that now, in this era, technology companies need to change their messaging strategy? Do you think there needs to be a major overhaul in terms of how the markets translate what they're actually trying to portray? My feeling is something needs to give.
Raphaelle d'Ornano
>> I think Wall Street, very rightly so, believes in the numbers. I think you need to have a coherent strategy. You need to have the numbers that can be operational and financial. The financial reality takes several quarters to actually translate, but there's a way of showing what road you are on with the right operational KPIs. You need to show. Again, the road you are on, the financial impact on revenue, on margins, on CapEx is never immediate, supposedly. So I don't think it's that they need to change the narrative. They need to show data points that investors can check, can believe, can benchmark, because this is a rational work. So I don't think it's more of a better narrative. It's more explaining what you do and giving the proof.
Gemma Allen
>> Sure. Speaking of messaging and masters of messaging, let's move to Sam Altman on OpenAI and their... I don't know how you could even put a phrase on their valuation, their 35X revenue valuation this week. Gearing up for an IPO, interesting time for that company too though, if we look at it in perspective. They had a pullback of Sora this week too, based on that mishap, I guess you could call it with Disney. They have also had some question marks over some of their behaviors broadly from the perspective of customer loyalty, et cetera. When we think about the Anthropic, Dario, Department of Defense debacle. So I feel like it's a company at an interesting point, but the valuation, it's astronomical. What are your thoughts? Do you think it's warranted? Where's your head at when you see these numbers?
Raphaelle d'Ornano
>> So I would say, first, from a coding perspective, their coding model is actually delivering outstanding performance. So I must give that point to OpenAI. In the past weeks, you see the pace of change, there has been a change in how their solution is actually, in many cases, actually preferred to Claude Code. This is very new. Now this leak could completely re-change the picture once again. But we've seen literally in the last not months, weeks, a change. So I think OpenAI is very clear that they are now hopefully starting to focus on, "How do we make money? How do we monetize our technology? No one is contesting the technology. No one has ever contested the actual technology, but how do we make money?" So indeed, Sora, all of those adventures, if we can call them like that, that is not where you find the most efficiency that's costing a lot of compute and that is maybe not the priority, if I may. What is the priority? How do you get this whole agentic AI underground through coding, through the enterprise, et cetera? So I think they're starting to go in that direction and they've recently said that they were shifting even their sales efforts, they're going all in into the enterprise and that's been a massive move. Again, this-
Gemma Allen
>> There has also been speculation about an ad model though, right? We saw the Super Bowl ads this year. I know they haven't actually probably really confirmed or denied that, but they haven't denied it either. And that would be a very different... If you think about it, then we're 10 years back again from the perspective of building a very proprietary data moat, like a Meta model all over again, right?
Raphaelle d'Ornano
>> Yes.
Gemma Allen
>> Versus an enterprise model where it's completely vertically integrated. Do you think you can hang your hat on both at the same time? Do you think that's a possibility for them?
Raphaelle d'Ornano
>> That's what they're doing for sure. They have a consumer play. I personally don't think that that is where they have a moat. They have a play, so they can go on that direction. It's what they've been doing with ChatGPT obviously. I personally don't think that there's a moat, that the moat is in the enterprise and in orchestration. They need to focus. When you're burning that level of cash, you need to focus. So you can maybe play two or three different things, not 10, so they're starting to realize that. The round was very difficult to raise. It was not an easy round to raise.
Gemma Allen
>> I actually did not know that.
Raphaelle d'Ornano
>> Oh, it was a super difficult round to raise. And I think investors know this in the background. There's been a lot of jokes of how you could very easily access the round if you wanted to.
Gemma Allen
>> Wow.
Raphaelle d'Ornano
>> And there's been a lot of jokes going in the background.
Gemma Allen
>> Do you think that there's a perception now they're on a complete moonshot? Do you think that is the broad perception of OpenAI in terms of the money, what they're looking for in terms of the cash that they're hemorrhaging?
Raphaelle d'Ornano
>> What I think is that we need to think about the impact that xAI, Anthropic and OpenAI, with their cumulative value, what they are going to represent for the IPO markets in the next months, not in the next years. I think that we need to imagine that a lot of these funds are going to be indexed. There's a big debate around that, that Americans and retirements are going to be having OpenAI, Anthropic, xAI shares. And I think that's the bigger question that needs to be asked because, if you have a model that is not clear and that is burning that much cash, maybe a regular American doesn't want that in his retirement account. And we're starting to see a lot of questions around that, which I think are the right questions because those three companies, given the size of their IPO, it's going to be very, very major.
Gemma Allen
>> Let's finish with one quick comment. I want to ask you about one quick thing. An interesting thing I did notice from this race is that there was a number of private investors through Arc Capital, which are... I guess they're probably very, very wealthy individuals, but they're not institutional investors. What are your thoughts on that? Does that signal anything to you? And do you think, to people listening, your thinking, are we going to see more and more of that? Are we going to see more of almost a retail investment thesis encroaching into the space? Because these company valuations are so huge, is the opportunity more democratized? What are your thoughts?
Raphaelle d'Ornano
>> Well, I think it's very dangerous. I think this is exactly the point. Look, I love AI. I obsess on AI. That's what I do all day long. I think those are, if you take Anthropic, OpenAI, xAI, MiniMax, DeepSeq, those are... And I keep repeating this... completely different animals with completely different readings of where is their mode, what does their P&L look like? MiniMax, DeepSeq and OpenAI are three very different species. So I think that it's already complex when you're doing that every day, every minute and literally obsessing over this, that when you're a regular American, it's not possible to have the right information. So those companies are generational companies. I agree. Anthropic, for me, could be the next Google or and I've been sharing these stakes a lot now. I think that it's important to have access, yes, but we're going to have to control this really in the right way because it's not just like, "Oh, let's all go to Disneyland and this is super cool."
No, no, this is not it. This is a very risky, still money-losing business with a lot of hard constraints, like power, capital, hyperscalers spending four trillion until 2030, if you look at the latest numbers. This is a dangerous game. Very attractive, yes. Very dangerous. So we should not be playing with the matches, if that's the expression, right?
Gemma Allen
>> Well, listen, Raphaelle, I know you have to go and we have a very busy day behind us here. This room is filling up with lots of very tech enthusiasts.
Raphaelle d'Ornano
>> .
Gemma Allen
>> The best and the brightest geeks of tech are here, I think today. So excited to talk to some of those folks. Thanks so much for coming on. I'm sure in two more weeks, I know you're going to France, you'll be back with us in two weeks. God knows what will have happened between now and then, but always love chatting with you.
Raphaelle d'Ornano
>> Thank you. Thank you.
Gemma Allen
>> Thanks so much for watching AGNT. Tune in next time.
