The Agent Wars Begin

Sam Hinton: Microsoft just announced they’re basically killing Copilot and I think this is the best thing that could happen to everyday users.

Alex Shannon: Wait, what? Microsoft is ditching their flagship AI assistant that they’ve been pushing for two years and you think that’s good news? You have thirty seconds to justify that.

Sam Hinton: Because Copilot was never supposed to be the endgame! It’s been a glorified chatbot with some integrations. What they’re building with Nvidia - actual autonomous agents - that’s the real revolution. Instead of asking Copilot to help you write an email, imagine an agent that just handles your entire workflow.

Alex Shannon: OK but we’re talking about replacing something people actually use with something completely theoretical. That sounds like a recipe for disaster.

Sam Hinton: Except it’s not theoretical anymore. OpenAI just released Codex with computer use capabilities - it can literally control Windows programs autonomously. The agent wars have officially begun.

Alex Shannon: Alright, now you have my attention. Let’s dive into this.

Alex Shannon: You’re listening to Build By AI, I’m Alex Shannon, and we are covering a day that might mark the official transition from AI assistants to AI agents.

Sam Hinton: And I’m Sam Hinton. Today we’ve got Microsoft and Nvidia teaming up to replace Copilot with actual autonomous agents, OpenAI solving decades-old math problems and taking control of Windows PCs, plus a massive funding round that has everyone scratching their heads.

Alex Shannon: We’ve also got SoftBank making a 75 billion euro bet on French data centers and GitHub developers in full revolt over new pricing. It’s May 31st, 2026.

Sam Hinton: And honestly, this feels like one of those days where we’re going to look back and say this is when everything changed. Let’s get into it.

Microsoft and Nvidia reportedly team up on AI PCs that run actual agents instead of Copilot

Alex Shannon: Alright, so let’s start with this Microsoft-Nvidia story that you were just teasing. According to reports from TechCrunch and The Decoder, Microsoft and Nvidia are collaborating to develop AI PCs that run autonomous agents rather than the traditional Copilot experience we’ve all gotten used to.

Alex Shannon: They’re planning to unveil new Windows computers from Dell and Microsoft Surface at Computex and Build, and the software is potentially based on something called OpenClaw. Sam, break this down for me - what’s the fundamental difference between what we have now and what they’re building?

Sam Hinton: OK so think about how you use Copilot today. You ask it a question, it gives you an answer, maybe it helps you write something. It’s reactive, right? You’re still doing all the actual work. These autonomous agents are proactive. They don’t wait for you to ask - they understand your goals and execute tasks independently.

Alex Shannon: But that sounds like a massive leap in complexity. How confident are we that they can actually deliver on this? Because we’ve seen a lot of AI promises that turned into disappointments.

Sam Hinton: That’s where the Nvidia partnership becomes crucial. Nvidia isn’t just providing chips here - they’re bringing their entire infrastructure for running complex AI models locally. And here’s the thing - we’re already seeing early versions of this working. Look at what OpenAI just shipped with Codex.

Alex Shannon: Right, but I’m thinking about the average user here. If I’m someone who just figured out how to use Copilot effectively, and now Microsoft is saying ‘forget all that, here’s something completely different,’ isn’t that going to create massive confusion?

Sam Hinton: You know what, that’s a fair point. But I think Microsoft learned from the Windows 8 disaster. They’re not going to just rip Copilot away overnight. This is probably going to be a gradual transition, maybe even opt-in at first.

Alex Shannon: The timing is interesting too, right? This comes at the same time that GitHub is changing Copilot’s pricing model and facing major backlash from developers. It feels like Microsoft is hedging their bets across their entire AI product line.

Sam Hinton: Exactly! And that’s smart strategy. Look, if autonomous agents work the way they’re promising, the whole concept of ‘AI assistants’ becomes obsolete. You don’t need an assistant if you have an agent that can actually accomplish tasks without you.

Alex Shannon: I guess the question is whether consumers are ready for that level of AI autonomy. There’s a big psychological difference between asking an AI for help and letting an AI just… do things on your behalf.

Sam Hinton: True, but I think we’re going to see adoption happen faster than people expect, especially if these agents can deliver on productivity gains. The early adopters will be businesses and power users who can immediately see the ROI.

Alex Shannon: Let me play devil’s advocate for a second though. What about trust and control? If an agent is making autonomous decisions about my work, my emails, my documents - what happens when it makes a mistake? With Copilot, I can review everything before it happens.

Sam Hinton: That’s exactly why the OpenClaw foundation is so interesting. From what we’re hearing, it’s designed to be transparent and auditable. You should be able to see what the agent is planning to do and intervene if needed.

Alex Shannon: But will regular users actually do that? Or will they just trust the agent to get things right? Because if it’s the latter, we could be setting up a lot of people for some really expensive mistakes.

Sam Hinton: Look, that’s always the trade-off with automation, right? The more automated it gets, the less control you have. But the productivity gains could be massive if they nail the implementation.

Alex Shannon: The Dell and Surface hardware announcements at Computex will be really telling. If they can show smooth, reliable agent operation in real-world scenarios, not just controlled demos, that could be the tipping point.

Sam Hinton: Absolutely. And honestly, if Microsoft and Nvidia can’t make this work with their combined resources and market position, it probably means the technology isn’t ready for mainstream adoption yet.

Alex Shannon: Keep an eye on those Computex and Build announcements. If Microsoft and Nvidia can actually demo this working smoothly, it’s going to put enormous pressure on Apple, Google, and everyone else to catch up quickly.

Sam Hinton: Which is exactly why this feels like such a pivotal moment. We’re not just talking about one company’s product roadmap - we’re talking about the next fundamental shift in how people interact with computers.

OpenAI’s AI model solves decades-old maths problem, proof verified by researchers

Alex Shannon: Speaking of OpenAI, let’s talk about what might be an even bigger story. Multiple sources are reporting that OpenAI’s AI model has solved a decades-old mathematics problem, and here’s the key part - the proof has been verified by researchers. This isn’t just AI generating something that looks right, this is mathematically sound work.

Sam Hinton: Dude, this is huge. We’re talking about moving from AI that can write code and essays to AI that can actually advance human knowledge in pure mathematics. That’s a fundamentally different category of capability.

Alex Shannon: But what does this actually mean in practical terms? I know this sounds impressive, but how does solving a decades-old math problem translate to real-world impact for most people?

Sam Hinton: OK so think about it this way - mathematics is the foundation of everything from encryption to engineering to physics. If AI can now solve problems that human mathematicians couldn’t crack for decades, that’s like giving every field that depends on math a massive upgrade.

Alex Shannon: That’s a good point, but I’m also thinking about the implications for academic research and scientific discovery more broadly. If AI can solve these kinds of theoretical problems, what happens to the role of human researchers?

Sam Hinton: I think it’s going to be more collaborative than competitive. The AI didn’t just magically solve this - it had to be guided and the proof had to be verified by human researchers. But now those researchers can tackle much bigger questions because they have this incredibly powerful tool.

Alex Shannon: Right, but there’s also this question of whether we can truly understand the solutions that AI generates. If an AI solves a math problem using methods that human mathematicians can’t easily follow, how do we verify that it’s actually correct beyond just checking the final answer?

Sam Hinton: That’s exactly why the verification by researchers is so important here. It suggests that the AI didn’t just brute force a solution - it generated work that human experts can examine, understand, and validate. That’s actually more impressive than if it had just spit out an answer.

Alex Shannon: I’m curious about the process here. Do we know anything about how long it took the AI to solve this problem compared to how long human mathematicians have been working on it?

Sam Hinton: The reports don’t give specific timelines, but we’re talking about a problem that’s been unsolved for decades being cracked by an AI system. Even if it took the AI weeks or months, that’s still an incredible acceleration of discovery.

Alex Shannon: And this ties back to what we were talking about with the Microsoft-Nvidia story, right? We’re seeing AI move from being assistive to being genuinely productive and creative in ways we hadn’t seen before.

Sam Hinton: Exactly! Whether it’s autonomous agents handling your workflow or AI solving mathematical proofs, we’re hitting this inflection point where AI isn’t just responding to human prompts - it’s actually advancing human capabilities in meaningful ways.

Alex Shannon: But here’s what worries me - if AI can solve decades-old math problems, what does that do to funding priorities in academic research? Why would you fund a team of human mathematicians to work on a problem when an AI might solve it in a fraction of the time?

Sam Hinton: That’s a legitimate concern. But I think the human element becomes even more important for defining which problems are worth solving and interpreting what the solutions actually mean in broader context.

Alex Shannon: The question is whether academic institutions and research organizations are prepared for this shift. If AI can solve decades-old problems, how do we restructure research priorities and funding?

Sam Hinton: I think we’re going to see a complete rethinking of how research gets done. Instead of spending years on problems that AI might solve in days, researchers can focus on defining the right problems and interpreting the solutions.

Alex Shannon: There’s also the question of reproducibility. If an AI solves a math problem, can another AI system independently verify that solution? Or do we still need human mathematicians as the ultimate arbiters of correctness?

Sam Hinton: Great question. I suspect we’ll need both - AI for speed and scale, humans for intuition and judgment. The combination could be incredibly powerful.

Alex Shannon: This is definitely something to watch closely. If OpenAI can replicate this success across other mathematical domains, we could be looking at an acceleration of scientific discovery unlike anything we’ve seen before.

Sam Hinton: And not just in pure math. This could revolutionize fields like cryptography, materials science, drug discovery - anywhere complex mathematical modeling is the bottleneck to progress.

OpenAI’s Codex can now operate your Windows PC autonomously, hunting bugs and testing apps on its own

Alex Shannon: And here’s where OpenAI’s advances get really practical. Reports from The Decoder and Google News indicate that OpenAI’s Codex application now runs on Windows 11 with something called ‘Computer Use’ capability. This allows the AI to autonomously control programs, test applications, and identify bugs.

Alex Shannon: But here’s what caught my attention - users can also remotely start and monitor tasks from the ChatGPT mobile app. So you could literally be on your phone, tell your AI to run tests on your desktop, and monitor the progress remotely.

Sam Hinton: OK this is exactly what I was talking about with the agent revolution. This isn’t an AI helping you write code - this is an AI actually using your computer like a human would. It can click buttons, open programs, run tests, find bugs. That’s autonomous operation.

Alex Shannon: But I have to ask - are we comfortable with AI having that level of control over our computers? There’s something that feels both exciting and terrifying about letting an AI loose on my desktop.

Sam Hinton: Yeah, that’s a legitimate concern. But think about who this is targeting first - developers and testers who are already running automated scripts and tools. For them, this is just a much more sophisticated version of automation they’re already using.

Alex Shannon: That makes sense, but what about security implications? If an AI can control programs and access files autonomously, what happens if something goes wrong? Or if the AI gets confused about what it’s supposed to be doing?

Sam Hinton: That’s where the remote monitoring through ChatGPT becomes crucial. You’re not just setting it loose and walking away - you can watch what it’s doing in real-time and intervene if needed. It’s autonomous but supervised.

Alex Shannon: I’m also thinking about the productivity implications here. If AI can handle testing and bug hunting autonomously, that could free up developers to focus on more creative and strategic work. But it could also make certain testing roles obsolete.

Sam Hinton: Absolutely, but I think we’ve seen this pattern before with other forms of automation. The roles don’t disappear - they evolve. Instead of manually testing applications, you might be designing test strategies and interpreting AI-generated test results.

Alex Shannon: Let me ask you this though - have you ever watched an automated testing suite go haywire and start clicking random things or getting stuck in loops? What makes us think an AI agent won’t have similar issues but potentially worse consequences?

Sam Hinton: Fair point! But that’s exactly why the supervised approach is so important. Traditional automation follows rigid scripts. This is supposed to be intelligent enough to adapt when things don’t go as planned.

Alex Shannon: Supposed to be is the key phrase there. I’m curious about error handling. When this AI agent encounters an unexpected dialog box or system error, does it know to stop and ask for help, or does it try to power through and potentially cause more problems?

Sam Hinton: That’s going to be the make-or-break factor for widespread adoption. If it can gracefully handle edge cases and know its limitations, this could be game-changing. If it can’t, it’ll be relegated to very controlled use cases.

Alex Shannon: The interesting thing is how this connects to the Microsoft-Nvidia story. If Microsoft is moving toward autonomous agents and OpenAI is already shipping autonomous computer control, we’re seeing the ecosystem align around this agent-based approach.

Sam Hinton: Right! And that’s not a coincidence. Microsoft has a significant relationship with OpenAI, so they’re probably getting early access to these capabilities and building their agent strategy around them.

Alex Shannon: It also raises questions about data security. If an AI agent is autonomously controlling your computer, what data is it accessing? What’s being logged? Where is that information going?

Sam Hinton: Those are crucial questions that I hope OpenAI is addressing transparently. For enterprise adoption, companies are going to need iron-clad guarantees about data handling and privacy.

Alex Shannon: And what about liability? If an AI agent autonomously deletes important files or corrupts data while hunting for bugs, who’s responsible? The user who initiated the task? OpenAI? The software vendor?

Sam Hinton: Those legal frameworks are definitely going to need to evolve quickly. We’re moving into uncharted territory where AI isn’t just generating content - it’s taking actions with real consequences.

Alex Shannon: For developers listening, this might be worth experimenting with now, especially if you’re doing repetitive testing tasks. The learning curve is probably going to be steep, but early adopters could see significant productivity gains.

Sam Hinton: And honestly, if you’re not at least exploring this kind of automation, you might find yourself at a competitive disadvantage pretty quickly. This feels like one of those technologies that’s going to become table stakes within a year or two.

Alex Shannon: Just make sure you’re doing it in a safe environment first. Maybe set up a virtual machine for the AI to play in before letting it loose on your main development setup.

‘What a joke’: Github Copilot’s new token-based billing spurs consternation among devs

Alex Shannon: Now let’s talk about something that’s got developers pretty fired up. Both TechCrunch and The Decoder are reporting that GitHub Copilot’s new token-based billing model has generated significant criticism among developers. The feedback has been harsh - we’re seeing reactions like ‘what a joke’ from the developer community.

Alex Shannon: This marks the end of what people are calling the free or low-cost period for the service. Sam, help me understand the anger here - is this just about price increases, or is there something deeper going on?

Sam Hinton: It’s definitely deeper than just pricing. Developers got used to Copilot being this relatively affordable, flat-rate tool that they could use freely. Moving to token-based billing makes the cost unpredictable and potentially much higher for heavy users.

Alex Shannon: But isn’t token-based billing pretty standard for AI services at this point? OpenAI, Anthropic, most of the major players use some version of usage-based pricing.

Sam Hinton: True, but those are API services for businesses. GitHub Copilot positioned itself as a developer productivity tool - more like an IDE plugin than an AI service. Imagine if your code editor suddenly started charging you per keystroke. That’s how this feels to developers.

Alex Shannon: That’s a great analogy. And the timing is interesting too, right? This happens just as Microsoft is announcing they’re moving away from Copilot toward autonomous agents. It feels like they’re squeezing revenue out of the old model while transitioning to something new.

Sam Hinton: Exactly! And I think that’s what’s really frustrating developers. They feel like they’re being asked to pay premium prices for a product that Microsoft is basically admitting is obsolete.

Alex Shannon: But playing devil’s advocate here - if Microsoft is investing heavily in autonomous agents and next-generation AI capabilities, don’t they need to fund that development somehow? Maybe this pricing change is necessary to support the innovation.

Sam Hinton: Look, I get the business logic, but the execution is terrible. Instead of just jacking up prices, they could have offered a migration path to the new agent-based tools or grandfathered existing users. This feels like they’re prioritizing short-term revenue over developer relationships.

Alex Shannon: And developer relationships are crucial for Microsoft, especially given their competition with Google, Amazon, and others in the cloud space. Alienating the developer community over pricing seems like a risky move.

Sam Hinton: Right, and here’s the thing - developers are already exploring alternatives. There are open source coding assistants, other commercial options. Microsoft built up this huge user base and now they’re giving people a reason to look elsewhere.

Alex Shannon: What’s particularly frustrating for developers is the unpredictability. With flat-rate pricing, you know exactly what your monthly bill will be. With token-based billing, your costs could vary wildly depending on how much you use the service.

Sam Hinton: And that’s especially problematic for freelancers and small teams who need to budget carefully. If you can’t predict your Copilot costs, it becomes much harder to factor into project pricing and budgets.

Alex Shannon: It also raises questions about the broader sustainability of AI-powered developer tools. If the underlying AI models are getting more expensive to run, are we going to see price increases across all these services?

Sam Hinton: Probably, yeah. The free lunch period for AI tools is definitely ending. But the companies that handle this transition well - by providing clear value and reasonable pricing - are going to win in the long run.

Alex Shannon: I wonder if this is also about Microsoft trying to segment their market. Maybe they want to push casual users toward their new autonomous agents while keeping Copilot as a premium service for power users.

Sam Hinton: That’s possible, but if that’s the strategy, they’re not communicating it well. Developers just see a price increase without a clear explanation of what they’re getting in return.

Alex Shannon: The timing with the agent announcements is really unfortunate. If Microsoft had launched autonomous agents first and then said ‘here’s our legacy Copilot pricing,’ developers might have been more understanding.

Sam Hinton: Exactly. Instead, it feels like they’re charging more for an inferior product while promising something better in the future. That’s not a great value proposition.

Alex Shannon: For developers who are unhappy with this change, now might be a good time to evaluate alternatives and diversify the tools they’re using. Don’t get too dependent on any single AI coding assistant.

Sam Hinton: Absolutely. And honestly, this might be a wake-up call for the entire ecosystem. If one provider can suddenly change their pricing model, you want to have backup options ready to go.

Alex Shannon: The silver lining is that this controversy might accelerate development of open source alternatives. Nothing motivates open source development like proprietary software getting expensive.

SoftBank says it will invest up to €75 billion to build French data centers

Alex Shannon: Alright, let’s hit some rapid fire stories. First up, early reports suggest that SoftBank announced plans to invest up to 75 billion euros in building French data centers, aiming to develop up to 5 gigawatts of additional capacity.

Sam Hinton: That’s an absolutely massive investment - we’re talking about enough capacity to power several major cities worth of computing. This screams AI infrastructure to me. SoftBank is betting huge on European AI demand.

Alex Shannon: The France location is interesting too. With all the EU AI regulations, having data centers in Europe is going to be crucial for compliance. SoftBank might be positioning themselves as the infrastructure backbone for European AI companies.

Sam Hinton: Exactly, and 5 gigawatts is enough to run some seriously large AI training operations. This could be what enables European companies to compete with US and Chinese AI development at scale.

Alex Shannon: I’m also wondering about the timeline here. 75 billion euros doesn’t get spent overnight - this is probably a multi-year buildout. But the fact that they’re committing this much capital suggests they see massive demand coming.

Sam Hinton: And SoftBank has a pretty good track record of spotting tech trends early, even if they don’t always nail the timing. This feels like them positioning for the AI infrastructure boom before it really hits Europe.

Alex Shannon: The energy requirements alone are staggering. 5 gigawatts of data center capacity is going to need serious power infrastructure. This could drive renewable energy development in France too.

Sam Hinton: True, and that might be part of the appeal for the French government. This isn’t just about data centers - it’s about positioning France as a major player in the global AI economy.

OpenRouter raises $113M Series B

Alex Shannon: Next, early reports from Hacker News suggest that OpenRouter has raised 113 million dollars in Series B funding, though details are still limited.

Sam Hinton: OpenRouter is that API platform that lets developers access multiple AI models through a single interface, right? A 113 million Series B suggests investors are betting big on the multi-model approach rather than being locked into a single provider.

Alex Shannon: That makes sense given what we’re seeing with pricing changes at individual providers. Having a platform that can switch between different AI models could become really valuable as the market matures.

Sam Hinton: Yeah, and with all these autonomous agents we’ve been talking about, you probably want access to specialized models for different tasks rather than trying to use one model for everything.

Alex Shannon: The timing is perfect too. As companies get burned by unexpected pricing changes from major providers, having a platform that offers choice and flexibility becomes more attractive.

Sam Hinton: Exactly. OpenRouter essentially provides insurance against vendor lock-in. If one AI provider jacks up their prices or changes their terms, you can switch to another model without rewriting your entire application.

Alex Shannon: 113 million is also enough to build serious enterprise features - things like guaranteed uptime, dedicated support, custom model fine-tuning. That could help them compete with the big players.

Sam Hinton: And they’re probably going to need those enterprise features as more businesses look for alternatives to direct relationships with OpenAI, Anthropic, and the other major providers.

Google’s Agentic AI Tool Gemini Spark Is Now Available - PCMag

Alex Shannon: Speaking of agents, early reports suggest that Google’s agentic AI tool Gemini Spark is now available to users, marking Google’s expansion into agent-based AI applications.

Sam Hinton: OK so now we’ve got Microsoft with autonomous agents, OpenAI with computer use, and Google with Gemini Spark. The agent wars are definitely heating up. Everyone’s trying to move beyond chatbots to actual task execution.

Alex Shannon: Google’s been playing catch-up in the AI space, so launching an agent platform makes sense. But they’ll need to differentiate somehow - just being another AI agent isn’t going to cut it.

Sam Hinton: Google’s advantage could be integration with their entire ecosystem - Gmail, Docs, Calendar, Search. An agent that can seamlessly work across all Google services could be pretty compelling for business users.

Alex Shannon: That’s true, especially for companies that are already invested in Google Workspace. If Gemini Spark can actually automate workflows across Gmail, Docs, and Sheets, that could be a major productivity boost.

Sam Hinton: The question is whether Google can execute on that vision. They have all the pieces in place, but they need to make them work together seamlessly. That’s harder than it sounds.

Alex Shannon: And the competitive pressure is intense. Microsoft is already working with Nvidia on hardware-accelerated agents, OpenAI has computer control working. Google can’t afford to be late to this party.

Sam Hinton: Right, and Google has struggled with AI product launches before. They need Gemini Spark to work flawlessly from day one, or they risk falling even further behind in the agent race.

Anthropic’s $65B Raise: Can Claude’s Enterprise Surge Justify a $965B Valuation? - The Futurum Group

Alex Shannon: And finally, early reports suggest that Anthropic has raised 65 billion dollars in funding, achieving a 965 billion dollar valuation. The question being raised is whether Claude’s enterprise growth can justify that massive valuation.

Sam Hinton: Wait, 965 billion? That would make Anthropic worth almost as much as Apple. I’m going to need to see some serious revenue numbers to justify that kind of valuation. That seems… aggressive.

Alex Shannon: Yeah, that valuation definitely raises eyebrows. Even with Claude’s enterprise success, we’re talking about a company that’s still primarily competing on being a better chatbot. The fundamentals would have to be incredible.

Sam Hinton: Unless they’ve got some major technological breakthrough we haven’t heard about yet, this feels like we might be hitting peak AI hype in terms of valuations. Those numbers are going to need to be backed up by real performance.

Alex Shannon: The enterprise angle is interesting though. If Anthropic has cracked the code on enterprise AI deployment at scale, that could justify premium valuations. Enterprise customers pay a lot for AI that actually works reliably.

Sam Hinton: True, but nearly a trillion dollar valuation? That suggests they’d need to capture a huge chunk of the entire enterprise software market, not just AI. The expectations are incredibly high.

Alex Shannon: This also puts enormous pressure on them to deliver. When you’re valued like the world’s most valuable company, anything less than extraordinary results is going to disappoint investors massively.

Sam Hinton: Exactly. And it makes me wonder about the sustainability of these AI valuations across the board. If Anthropic is worth nearly a trillion, what does that imply about OpenAI, Google, Microsoft? The numbers start getting pretty wild.

BIGGER PICTURE

Alex Shannon: Alright, if you zoom out and look at everything we covered today, there’s a really clear theme emerging. We’re seeing this massive shift from AI assistants that help you do things to AI agents that actually do things for you.

Sam Hinton: Right, and it’s happening across the board. Microsoft is ditching Copilot for autonomous agents, OpenAI is shipping computer control, Google is launching agent tools. Even the infrastructure investments and funding rounds are focused on supporting this agent-based future.

Alex Shannon: But I think the GitHub Copilot pricing backlash is a warning sign. As companies transition from the current generation of AI tools to agents, how they handle that transition is going to make or break user adoption.

Sam Hinton: Absolutely. And the winners are going to be the companies that can deliver genuine autonomous capability, not just rebranded chatbots with better marketing. The bar for what counts as ‘AI agent’ is about to get much higher.

Alex Shannon: The infrastructure story is fascinating too. SoftBank dropping 75 billion euros on French data centers isn’t just about capacity - it’s about positioning Europe to compete in this agent-based future. That level of investment suggests they see massive demand coming.

Sam Hinton: And that connects to the OpenRouter funding round too. As we move toward more complex agent workflows, you’re going to need platforms that can orchestrate multiple AI models and services. The single-provider approach might not be sufficient anymore.

Alex Shannon: What’s interesting is how the mathematical breakthrough fits into this pattern. OpenAI solving decades-old math problems shows that AI is moving beyond just automating existing human tasks - it’s actually advancing human knowledge in fundamental ways.

Sam Hinton: Exactly! And when you combine that level of problem-solving capability with computer control and autonomous operation, you get agents that aren’t just following scripts - they’re genuinely intelligent workers that can adapt and discover.

Alex Shannon: But there’s also this tension between the promise and the current reality. Anthropic’s trillion-dollar valuation suggests massive investor confidence, but the GitHub pricing backlash shows that users aren’t necessarily ready to pay premium prices for current-generation AI tools.

Sam Hinton: That’s exactly the gap that needs to be bridged. The technology is advancing incredibly quickly, but user adoption and willingness to pay is going to depend on delivering real, measurable value - not just impressive demos.

Alex Shannon: And the competitive dynamics are getting really interesting. We’re seeing companies like Microsoft make massive strategic pivots, essentially abandoning products they’ve spent years building in favor of next-generation approaches.

Sam Hinton: Right, and that’s both exciting and risky. If autonomous agents deliver on their promise, Microsoft looks prescient. If they don’t, they’ve alienated their user base and handed market share to competitors.

Alex Shannon: What should people be watching for over the next few months? What signals will tell us whether this agent revolution is real or just hype?

Sam Hinton: Look for real demonstrations of complex task completion. Not demos where an AI writes an email, but where it actually completes multi-step workflows autonomously. And watch the enterprise adoption numbers - businesses will be the first to pay premium prices for genuine productivity gains.

Alex Shannon: I think we’re also going to see a shakeout in terms of which companies can actually deliver on these agent promises versus which ones are just riding the hype wave. The valuations we’re seeing today won’t survive if the technology doesn’t deliver.

Sam Hinton: And pay attention to the infrastructure buildout. If companies are really preparing for agent-scale computing demand, we should see massive investments in data centers, specialized chips, and networking capacity. SoftBank’s French investment is just the beginning.

Alex Shannon: The regulatory piece is going to be crucial too. As AI agents become more autonomous and capable, governments are going to need to figure out liability, safety, and oversight frameworks. That could accelerate or derail adoption depending on how it’s handled.

Sam Hinton: True, and we’re already seeing that with the EU’s AI regulations. Companies that can navigate those regulatory requirements while delivering genuine agent capabilities are going to have a huge competitive advantage.

OUTRO

Sam Hinton: This has been Build By AI. If you’re as fascinated by this agent revolution as we are, make sure you subscribe so you don’t miss how this plays out.

Alex Shannon: Tomorrow we’ll be covering whatever new developments emerge from this rapidly evolving space. The pace of change is honestly incredible right now.

Sam Hinton: Thanks for listening, and we’ll see you tomorrow.