In this episode of the Future of Distributed Engineering (FDE) podcast, hosts A and B discuss how compute has become a core competitive advantage across multiple AI‑related domains. They begin by noting AI’s breakout into mainstream advertising, citing the Super Bowl ads from OpenAI, Anthropic and Amazon as evidence that AI is moving from a niche technology to a brand‑level infrastructure. The conversation then shifts to SpaceX’s moon‑city vision and a forecast that 2026 could be a massive IPO year for the company, framing lunar logistics as the next industrial frontier. They explore the emergence of automated scientific discovery, where AI plans experiments and robots execute them, while acknowledging current limits in robotic dexterity. The hosts argue that AI is now “eating atoms,” describing how advanced materials for AI substrates are creating supply‑chain constraints and tying AI hardware costs to corporate cash‑flow pressures. Robotics is highlighted as leaving the lab, with real‑world deployments such as robots handling snow and Tesla’s high‑volume production of electric semitrucks. A discussion on labor market polarization follows, noting a split between super‑users who leverage AI for massive productivity gains and a broader workforce facing compression, and warning of a winner‑take‑all dynamic. The episode then delves into Claude Cowork, the speaker’s personal adoption of the tool, its plans, and how it replaces many traditional workflows, positioning it as an enterprise AI co‑pilot. Finally, they address the growing AI‑agent ecosystem, the addition of VirusTotal scanning for security, and the risk that AI agents may become an app‑store‑like marketplace with heightened security challenges. Throughout, the hosts blend conversational anecdotes with analytical observations, offering a snapshot of how compute, AI hardware, and software are reshaping technology, industry, and work.
Chapters
The spaces. It's our 13th episode so far. So the theme of this week is that AI is no longer something that is quietly happening in tech circles. Only our first news is related to that. AI goes mainstream. What happened basically this year Superball ads quietly become an AI showcase. OpenAI run a ad related to the codecs that you can build. You can just build things at the same time. Anthropic position itself that no ads in AI conversations. And Amazon jumped into that as well that Amazon lead into Alexa plus AI at home. That was a bit funny, useful and scary at the same time. So what's your take on it? Super fun.
So Superball is a proxy for mass adoption and historically there were Cokes and Pepsis and Apple and all that. But this year I've seen like it was all AI battlefield. So at one point it was Codex and they were branding like you cannot just. So Codex is not just OpenAI is not just chat and now you can build and ship things from it. Like and on the other hand Anthropic was saying like you can use us without any ad. So of course you see that both Anthropic and OpenAI they're fighting on this platform. So it's fun to see that how AI has become. This is a platform for this much mass adoption. So yeah, I mean it's, it's. I think, I think if we have to see like if we zoom out. So we are moving from AI as a novelty to AI as a brand and AI the default infrastructure. So it's, it's a, it's a big shift. So business and also I mean Business Insider covered how Anthropic Superball directly took a shot at OpenAI's direction. And the broader pattern is that AI is now mainstream advertising. So now people are using this as a news report sharing for their own platforms. Like how these two big companies are fighting on the super bowl ad space.
Truly. So moving on to the next news of the day is related to the SpaceX the moon and IPO gravity. So what happened? Basically even those Elon tweeted today as well that and emphasized a vision of self growing cities on Moon within 10 years and argued that moon is faster to industrialize than Mars. Even though there's a report that is being shared that 2026 would be a major IPO year for the SpaceX floating and a potential and predicted to trillion dollar companies as in whole. And what do you think if SpaceX build a system that allow anyone to travel to the moon? How fascinating and how industrial shift that's going to be for industry as well.
So first of all, I think anything significantly advanced in technology is indistinguishable from magic. Arthur C. Clarke is a sci fi writer. Right. So if you ask, if you go into, I mean what we are saying like there would be a self grown city or people would be traveling to moon, it's equivalent to 16th century where somebody's saying probably would be traveling the world and all that. We'll be flying around the world and all that stuff. So I think if we need to understand like the things that we have now is. I mean if someone from the 16th century come at this point in time, they would be just surprised what's happening there. It's already feel like magic. So as Moon. So I think what's happening here is like for all those years we built that all that infrastructure to travel to Mars and figure out probably it's. It's not as close as we think in terms of building cities there and going and all that. But Moon is. I mean we can literally build cities in the moon. And there are a couple of steps here. The first is like in next two to three years if we were successfully able to build that orbital infrastructure that's required to power AI, the next would be Moon. We could easily. I mean in the next 10 years it won't feel like as magic that really can we travel to moon. So that's what Elon is talking about. So moon pivot is more of a logistics story.
So.
So faster cycles, closer adaption, closer distance, more iterations. And if you're serious about civilizational insurance, I think iteration beats romance. So that's this moon story all about so far.
So science becomes a robot plus AI. There are leaders like Kevin Vale argue that there's no reason for human to perform repetitive labor work where AI can plan things for us and robots can execute them for us, creating a very close loop of discovery. So we are already entered into an era of automated discovery. But do you think who going to breakfast? Universities, pharma or energy all.
No, I think what's happening. Have you, have you done your. I mean when we were studying have you done some labs like how we used to do labs like it's. We don't go to the physics labs. How boring they were. And they were all same stuff we used to take readings and all that. So that Kevin is arguing that AI would reason what to do. Clearly they're very good in reasoning and. And then robots would take that reasoning and perform the physical experiment. So we don't need grad students. What they are going to do that. So clearly you are. What you are saying is like you are just outsourcing the process of discovery to the robots and AI. I'm not sure how far along we are because we, I mean if we have to automate, I mean reasoning is good, but if we have to automate the work of physical labor that's required in the lab. So we need a very high quality hand dexterity that's currently not available in the robot's at stage. But if we have that maybe two, three years down the line then only it's not like just the lab would be automated. I would have to find a work which won't be automated. Whether we do that kind of work with robots and combine it with AI.
Got it. So we used to talk about that AI is eating software, but it's like now AI is eating atom, not just software. What happened? Basically there are advanced materials that are like glass block. They are being used in AI substrates. So they are becoming supplier constraint. At the same time, Amazon Free cash flow pressures is being directly linked to the massive AI capsules. So it's not all of a sudden, but we are looking at the shortage of supply chain for the AI and AI is becoming more of infrastructure story. Now that's not just related to one.
It was all infrastructure story, not even just AI. Software was an. Code was an infrastructure story. So Abhissan wrote this book and the book I think third or fourth page there was like exact sentence code can find root in any physical substrate it's running on. So what it really means is like if you mutate the physical world it. It generates code on top of it, right? So think about like in your brain, billions of neurons are firing. It's an entirely physical phenomena. The thought is a purely abstract thing. But can you think of like thought would happen without firing of neurons or neurons would fire and no thought would happen. So it's already linked. So it's just like now with AI we're just feeling this like how these data centers are creating this massive intelligence. They were being. At the end of the day, when you run the controlled flow of energy in any substrate and you mutate it, it would generate certain form of patterns. Right. Previously to harness that pattern it was software, but now it's AI. So I think and it's all physical. It was never just on the air. I mean your thoughts are very much physical phenomena, but that runs on top of your neurons. Similarly it was always a physical phenomena. But yes, the kind of things we require to produce this level of intelligence. It requires lots of physical infrastructure. Like we need fiber optics, substrate transformers, cooling, power, land permits. All that thing is required to produce this level of intelligence. So that's why some of our people are changing the name of the class with Claude cowork and it took them 30 minutes or 25 minutes. Asked like probably if you know how hard it is to do all this just to change the name of the class, probably you would do it yourself. But that's how convenience people would go for convenience and pay for it as much as much as required. So convenience is a big thing. So anyway, so it's all. So the thing is my favorite topic by the way. It's that AI for the physical infrastructure or both software and hardware was always related. It wasn't thinking the silos. It was always the same thing. Just like your neuron firing, your thoughts are the same thing.
Exactly. So moving on that robotics leaving the labs now. They are not confined to only clean demo environment. They are footage is being circulated from a unitary humanide. Reports attempted to clear the snow. They are being slipping, adjusting and even reacting to unpredictable conditions as well. At the same time there are some multiple reports that are indicating that Tesla electric semitrex is entering the high production this year. Like the massive production is going for their volume. These are the things that indicates to us that robotics are more coming towards the real world. We are looking towards these trends but what's your take on it?
How the world gonna be first for snow? Like yeah, I think the snow thing matters but it's a reality. It's a low traction, uncertain services, random edge cases. So robotics here is not intelligence robustness. So that's, that's a good surface area to try how robotics perform and for the, for. For the manufacturing for the semis. I think high volume manufacturing is when a product becomes an industry and that's from cool to economic force. So with semis. Yes, we're seeing this.
Exactly. So moving on to the next news is that labor is splitting super users versus everyone else. There are report from the Financial Times that highlight growing sentiments that enterprises, not consumer apps is where the more value is. AI is getting more values into the enterprise level instead of the consumer apps. So I did same time spent stories are emerging of individuals using AI to produce extreme output while the entry level white collar jobs are showing signs of compression at the same time. So what do you think? AI is creating a more winner take labor market and what should young people or generation needs to prepare for?
So There are multiple questions here and there are multiple things to talk about. It's a very important topic. So first of all, AI has always been any technology like AI is extremely centralized in its nature. So you would see it's always been a take all. So I would definitely be winner take all if not do anything about it. So there will be, you would see at the end there were three, four major players would be capturing most of the world in terms of value creation. And that should not happen as again I would recommend Amir Sands book like he calls it that why we should build technologies of freedom why we should build open source AI because at the end of the day there should be multiple winners. But that's one part that yes, I think there will be so many, there will be only few winners at the top. On the other hand, you talk about power users. Are there a few people who are using AI and, and that that are using AI in a way that AI is making them 200x productive or 20x productive? What is it? So yes, I think I've seen lots of people and including myself that all my latent ideas and I although I'm a code, I code but still there were so many things I wanted to do but because of the lack of the time or because I was involved in so many things I couldn't do. But now I'm doing it. So I'm spending previously was spending 12 hours in the work, so now I'm probably doing 15 hours. It's not like for me AI is a big left. That's one thing. And the only people and for the people who are not using AI beyond ChatGPT are the people who are basically who don't have so many latent ideas. So we're living in idea economy. So if you have lots of latent ideas. So for example, if you're in sales and you want to think about how can you be instantly in front of everyone if someone is thinking about a service or product you are selling, how you can achieve these kind of things, then you probably start thinking okay, what how AI can help? What are the things I can do? What are things I can't do? So you should have lots of ideas that you need to execute. Otherwise you will be just focusing on chargeability. On the pricing side. Yes, enterprise have money. So Claude has this very specific focus on the enterprise. It's not retail AI. ChatGPT was focusing on retail, but at the end of the day I think both would converge. So currently, previously Claude was just for enterprises who wanted to increase their Speed for coding. But now Claude Cowork is so good. So, so, so damn good. I mean, if you are using cloud cowork, you literally feel like you're using AGI. I'm telling you. Do you use cloud cover?
I've seen you using cloud cover a lot. I would like that you share your experience with us related to the cloud cover specifically and what kind of uses I do.
So I have this cloud cover and code. So with cloud I built apps. Cloud cover I use for lots of my sales stuff. Lots of. I try lots of different ideas. I use that lot of cloud cowork, then lots of coordination. So of course I run multiple teams. So there's so much. It's so hard to do all. I mean to run these different teams. But now cloud cover does it for me because it has access to my file system. You have to use it basically, by the way. But if you are using it securely, then it is doing emails for me, it is creating. I have to present a lot of. I meet with customers a lot. So I build apps, I do a code, I do lots of coding work. So Claude is doing all this for me and with all that still I have lots of free time where I think, okay, what next? What next? What next? So I think and I have Max plan and that's also the limit is also, I mean even in the max plan I have completed my limit. I'm not. I think like probably I should have more and it's 200. But even if you're not using Max plan, then there's another plan called it's a hundred dollars or then there's enough General Pro Plan $20. I think everyone should use Claude Cowork and think about ideas. You first need to think about what kind of ideas do you want to do. If it's an app, if it's a. If it's a software, every thing that you thought of, probably it's possible now to build with Claude or so my experience. So I've left ChatGPT by the way, previously I was using Pro, but I've just left it. I've left Grok, I've reduced everything to Claude.
So because you move more from informational things to beyond the executing part.
It's great thing. So we need to move from just so AI historically was very good from middle to middle. So you have to start it and then you have to end it in the middle was very good. So Claude cover actually now I think Codex and everyone would do it eventually. But Claude Cowork is basically it's not middle to middle. You have to start it. If you it's not like cloud bot. You have to, of course you want that level of control, but it starts and then it ends, right? So you just need to come up with great ideas.
So the last news is related to AI agents are becoming an ecosystem. So what's happening basically that an AI agent platform I just recently added virus total scanning for agent skills that follow your malware concerns, highlighting early security risk. Because that is getting really, really important in the agents marketplace. Separate these things from a playful experiment and allow agents to fully autonomously chat about this, about your things and observe as well. So there is basically question is that we agents are becoming ecosystem at the same time we're providing working with the security sides as well. So do you think that we are reinventing the app store with AI agents but significantly with higher security risks.
So first of all I think it's important to explain here what, what is skill? Maybe most of you don't know what is skill. So. So for example, if you're using cloud, which is cloud skills, it's basically how to guide your work. So imagine if you are basically if you are in sales and you have to help if you, if you want to help your salespeople how to price a certain deal. So you write a skill for that, right? You want your do to basically use that skill to price any day, right? Or for example if you're building a game for 5 year old or for 10 year olds. So you write a skill for that to guide someone how to build a game. So it's basically an sop if you like. It's a work guidance, right? Now you wrote it for humans, but now imagine this skill is accessible through a bot, right? And now everything that a human can do, a bot like these multis can do, right? Accessing those skills. So that is extremely powerful at one end, but it's very risky on the other end.
