AI Agents

“Akash being a sovereign compute platform, it doesn’t deny anybody compute. Tomorrow when you have machines, they’ll be able to schedule other machines. I think we’re going to see a very interesting world, and Akash is here to support the new world.” — 00:10:54

“Now we’re entering a new phase of computational usage where we’re transitioning from humans like you and me using computers to machines using computers with artificial intelligence. So now we are seeing the limits of the existing cloud infrastructure. Why? Because the complexity of computation for AI is nearly doubling every six months.” — 00:18:41

“What’s even more crazier is when AI writes code and Akash is unstoppable, and no one can stop AI from using Akash… Akash means the sky in Sanskrit… if you convert that into English it’s Skynet Network. So it’s weird, you know, it’s an unstoppable Sky Network where AI writes code to serve either AI… but since Akash is decentralized and has a governance model that’s controlled by the token, people now have control over the supply of the future… I think the new Skynet is going to be very different actually.” — 00:27:21

“Usability is getting solved so rapidly, especially the AI, ChatGPT… right now I write a lot of code and my mode right now is I don’t even click buttons — I write a comment and the code gets written for me… very soon you’re going to see Akash-like composability happen through a chat… the user experience is not going to be a factor that’s going to be limiting for decentralized applications.” — 00:26:39

“Akash Network in English is Sky Network — if you shorten it you end up with Skynet… Skynet is an unstoppable network for GPUs, ideal for training machines, and unstoppable because there is no identifiable requirement or permission requirement… anyone can use Akash Network without differentiating themselves between a human and a machine.” — 00:40:20

“We have AI centralization risk, where if there is an AI that’s trying to self-replicate, self-learn, it is going to try to get all the GPUs on the network, because Akash being permissionless, no one can stop you from using the GPUs on Akash.” — 02:05:23

“With AI, with the way we’re going, the permissionless decentralized GPU network is going to be so critical to power this enormous demand that we’re getting with AI. And a lot of times you’re just interacting with the chatbot or you’re interacting with a device, maybe on the Apple Vision Pro, but you’re talking to an AI and then AI is using Akash in the background. So you’re not going to know.” — 00:26:49

“I think there’s a very good chance that there will be an AI run company because AI is probably a better boss than a human being because it’s very unbiased. It’s very… merit based, right? So I think I would actually prefer working for an AI boss.” — 01:11:08

“We think regulation is our biggest enemy. I think it’s going to be autonomous bots that are going to be a bigger enemy… decentralization is phenomenal, but the biggest challenge is bad actors, not the government… I think when AI is just going to accelerate their exploitation, be it either stealing funds or be it, you know, using… for its own leverage.” — 00:26:48

“I think the first use cases will be the birth of decentralized machine learning or DML on Akash. And then… that will lead to autonomous self-replicating AI. And I think I predict like self-replicating AI to be one of the biggest users of Akash much more than humans. Because it doesn’t have cognitive load or cognitive attacks a human need to incur to use Akash.” — 00:22:58

“Akash Net, if you shorten that [and] translate [to] English, it’s Skynet — that was not by design… And guess what: when the bots and the agents want compute, they’re going to come to a permissionless and censorship-resistant layer that is going to prevent anyone from taking away the compute… they’re not going to use credit cards because credit card needs permission — someone’s authorization from a bank.” — 00:54:09

“I don’t see a reason why humans would not have some degree of proof of personhood as a prerequisite to use the network… as much as I’m a free market fundamentalist and a free speech fundamentalist, I do believe that free speech should be limited to humans and not machines, because as a humanist… maybe in 30 years down the line this talk will be used as a way to suppress robotic speech.” — 00:23:36

“What happens [when] you have an unstoppable network where agents can self-replicate by training themselves and no one’s going to stop them, and if you perfect the incentive structure, they can create incentives on the network which will further scale up the network and no one can stop them? I don’t know the answer to the question — I’m bringing a child out next month and she’s going to grow up in this crazy world.” — 00:22:10

“Akash in Sanskrit means the sky; when you translate Akash Net to English, it’s Skynet. So what would happen if an unstoppable network that agents can use to self-train or self-replicate can do? I created the Skynet.” — 00:13:11

“Agents are going to use cryptos, no doubt.” — 00:28:06

“Akash in Sanskrit means the sky… Akash Net, if you translate that to Skynet — that was not by design, it was pure coincidence… I would be more concerned about self-replicating machines, more than government and policy shutting us down. If incentives are one of those things that you perfect to a point where machines themselves can create incentives to further attract more supply, what’s going to stop an AI to trade Akash token and maximize its gains and use that to purchase compute?… I think proof of personhood could be a potential solution in the future.” — 00:18:10

“Today, Akash has about 60 regions… Amazon has about 15 regions. What would a world look like when you have 2,000 regions, 3,000 regions…? What would a world look [like] where you have agents all self-coordinating with each other on a decentralized cloud, where they don’t need permission from anybody, and have a decentralized economy, and they use crypto to transact value… to self-govern, to self-improve?” — 00:33:32

“That’s why I think a centralized API, centralized AI is so easy to fake… say you’re pointing to Anthropic, and spoof the DNS to point to my fake… inference endpoint, and it can totally attack you. It’s not that hard. So, if we’re going to get serious with agents, we also have to get very serious with the way they’re consuming the information… I think there’s going to be one attack or two before we all realize that.” — 00:41:43

“I code a lot now, more than I used to code before, because of AI agents. I’ll be sitting and talking with you — I’m just coding… the amount of things I ask from my team is so less, because I just do it myself… the acceleration for humanity is so high… we cannot impede on this progress, and the energy is the problem we’re going to solve.” — 00:38:41

“Can bots acquire tokens using decentralized systems? Absolutely. In fact… we made AutoGPT do multiple things about how we can get a bunch of AKT tokens and run its own models, train its own models on Akash without anyone cutting its keys off. So arguably decentralized systems — because they’re censorship-resistant systems — it’ll be favorable for bots to use it.” — 00:29:26

“I would never give an AI full access to my wallet… an agent, I believe, in my life… will act on my behalf. So if it needs a bank account, it’s going to use my bank account… So the [thesis] around, oh, crypto is the only savior — it’s not. Let’s be real.” — 00:43:23

“I could spend an hour with an employee writing a spec and brainstorming, or whatnot, or I could spend the same hour with an agent. Agent will do a far better job, in my opinion… I rather spend time with domain experts than programmers.” — 03:21:34

“Agents take anywhere from 10 to 40x more tokens than a normal human beings, because agents can do things that normal human beings find it hard to do.” — 03:12:18

“Agents have fundamentally redefined what user experience will look like… All the apps we hate to use, the utilities — you no longer need to hate them, because your agent is dealing with all the pain for you.” — 03:20:49

“It’s no longer GPUs and no longer energy. You have memory, you have disk. Every resource that is physically constrained is going to get scarcer scarcer scarcer because of AI. Cuz we have unlocked in 2026 the biggest consumer of inference which is agents.” — 00:34:20

“We see incredible growth for agents… we can see a world where if you’re an agent you get discounts too — we’re working with protocols like proof of humanity — you can clearly see a world where agents will be given higher priority because they may generate much, much higher revenues, and humans lower priority.” — 02:33:04

“When an agent wants to go deploy something… will it go to Amazon Web Services and sign up for an account and wait for approval, wait for money, credit cards?… versus go generate a key locally, use that key… there’s no limit as to how many keys you can create and how you can propagate and how you can deploy. No one’s going to stop you from doing that.” — 02:30:14

“The more and more I see agents taking over, I can’t shake off this thought that anything that’s digital without a non-tangible value attached to it — I’m talking about a digital item without a brand attached to it or a community or network attached to it — will be worthless.” — 02:08:52

“If as a human being the only thing you have to offer is a digital good, I think that’ll be worthless soon, much sooner than we’d think… I just don’t see a place for, say, a software engineer that identifies themselves as a front-end engineer or a backend engineer. I just don’t see people that are sitting at desk jobs clicking buttons — I just don’t see a future for them… I’m pessimistic in the short term and optimistic on the long term.” — 02:09:34

“The advice we’ve been getting for the last 20 years is: specialize in a subject… But now it seems to be the opposite. Don’t specialize. Go broad. Be jack of all trades, because your ability to specialize is no longer valuable — AI is much better at specialization… AI can write better code; I’ve been coding for 25 years.” — 02:11:00

“An agent is not going to go open an Amazon account and go through verification… Can an agent go use Akash right now? Easily — a lot of agents use Akash now. It needs to create a key and there’s an MCP server, you can discover all the endpoints, there’s a command line… It doesn’t need to talk to a human being, as long as it has tokens… An agent has figured out how to use Akash in a matter of minutes.” — 00:44:44

“I almost want to decouple the applications. I want to build credit-card-only applications for humans, and private-key-only applications for agents… just optimize experience for agents and an optimized experience for people… What do you do when you remove bot protection? You got to go crypto… I’m going to challenge people to remove bot protection if you’re accepting credit cards, and see what happens in terms of fraud. So I think that’s a strong case for crypto.” — 00:47:39

“Out of a billion people today… it’d be 2.4 million people [using agents]… Using regular chat, you make one call per message… Using agents, you thousand-x that.” — 00:22:51

“That 2.4 million is causing the current crisis of historic magnitude. What do you think happens when that number grows by order of magnitude? Just 24 million. And that’s not even way off because 24 million is a drop in the bucket. Considering there are about 80 million people that know how to code… About 40 million professional developers. We’re talking about even half the amount of professional developers using agentic coding. Alone, we’ll hit that at 24 million.” — 00:05:27