The AI Coding Revolution Has a Huge Problem?

The AI Coding Revolution Has a Huge Problem?

A few weeks ago, I stumbled across a debate that has been making the rounds in software engineering circles. The spark came from Boris Cherny, an engineer at Anthropic and the creator of Claude Code, arguably the most influential AI Agentic Coding harness in the world today. During a podcast appearance, Boris made a statement that immediately grabbed my attention: Coding is largely a solved problem. He went on to explain that he hadn’t written a line of code by hand since November, and that essentially all of his code is now authored by Claude Code. Needless to say, this generated strong reactions. Some people interpreted it as evidence that software engineering is about to be fully automated. Others saw it as confirmation that AI coding tools are delivering unprecedented productivity gains. As someone with twenty-five years of experience in software development, I found myself somewhere in the middle. Because while I absolutely believe AI is transforming software engineering, my own experiences suggest that programming is nowhere close to being a solved problem. In fact, the more I use these tools, the more complicated the situation appears. The Rise of Agentic Software Development Over the past few years, we’ve witnessed a rapid evolution in how software gets built. First, developers used AI to generate snippets of code. Then they began using AI assistants to complete larger programming tasks. Today, we’re entering what many people call Agentic Software Development. Instead of asking an AI for a few lines of code, developers increasingly delegate entire workflows. The AI can analyze requirements. Generate designs. Write code. Create tests. Review its own output. Deploy software. Monitor production systems. In theory, the human becomes less of a programmer and more of an orchestrator. The promise is obvious. If AI agents can perform most of the implementation work, then software engineers can become dramatically more productive. Ten times more productive, according to some advocates. Perhaps even more. At least, that’s the dream. My First Encounter With Enterprise Agentic AI In 2025, I returned to work after a lengthy medical leave. My wife had experienced a serious health crisis, and I had spent months focused almost entirely on family. When I came back, one of the first things I noticed was that my employer had become obsessed with Agentic AI. Leadership had heard about tools like Claude Code and Cursor. They had heard stories about developers becoming ten times more productive. Naturally, they concluded that our company needed its own internal version. Let’s call it Kevin. Kevin was our homegrown agentic harness. Compared to Claude Code, Kevin felt slower, heavier, and burdened with enterprise compliance guardrails. It ran on older-generation models. It often struggled with context. And yet, despite all its flaws, Kevin was still capable of orchestrating significant portions of software development. Using Kevin gave me a front-row seat to what Agentic AI actually looks like inside a large enterprise. What I observed left me both impressed and concerned. The Business Knowledge Problem One of the biggest weaknesses I encountered had nothing to do with coding itself. It had to do with understanding. AI models are remarkably good at generating software. What they are not particularly good at is understanding the business context behind that software. Organizations often operate on thousands of unwritten assumptions. Knowledge exists in hallway conversations. Slack threads. Meeting notes. Institutional memory. The heads of senior employees. Much of this information never appears in formal documentation. Humans navigate these gaps naturally. AI agents do not. If a requirement is not explicitly documented, the AI will often substitute something that appears reasonable based on its training data. The generated code may compile successfully. The unit tests may pass. The architecture may look elegant. And yet the implementation may completely miss the actual business objective. This problem becomes particularly severe in large brownfield systems where decades of accumulated business logic exist beneath the surface. The Context Window Wall Another challenge is context. Every AI model has a limited working memory. For small projects, this isn’t a major issue. For enterprise software systems containing millions of lines of code, it becomes a constant battle. Once an AI agent exceeds its effective context window, strange things begin to happen. The model forgets previous decisions. It hallucinates APIs. It reintroduces deprecated libraries. It generates solutions that were already rejected earlier in the workflow. Developers have created countless mitigation strategies. Context compaction. Summarization. Subagents. Selective context filtering. These techniques help. But they don’t eliminate the underlying limitation. The larger the system becomes, the harder it is for the AI to maintain a coherent mental model of the entire application. Ironically, this is often where software engineering is most difficult in the first place. The Hidden Cost of Infinite Code One thing AI agents are exceptionally good at is generating code. Lots of code. An astonishing amount of code. Thousands of lines. Tens of thousands of lines. Entire subsystems can appear almost instantly. The problem is that quantity and quality are not the same thing. Many AI-generated implementations contain unnecessary abstraction layers. Duplicate functionality. Excessive complexity. Architectural choices that seem reasonable locally but become problematic globally. Without strong human oversight, repositories begin accumulating what can only be described as AI sediment. Layer upon layer of generated code. Each piece understandable in isolation. Collectively becoming harder and harder to maintain. Technical debt compounds quietly. And unlike financial debt, software debt often remains invisible until it becomes a crisis. The Vibecoding Trap This brings us to what I believe is the most important problem. Human review capacity. A team of AI agents can generate thousands of lines of code per hour. A human engineer cannot review thousands of lines of code per hour with high confidence. The math simply doesn’t work. As output increases, review quality inevitably declines. Cognitive fatigue sets in. Attention drops. Comprehension weakens. Eventually, the reviewer stops acting as an engineer and starts acting as a rubber stamp. This is the danger behind what many people call vibecoding. The developer repeatedly prompts the AI until something appears to work. The code ships. Nobody fully understands it. Nobody feels ownership over it. And nobody wants to maintain it six months later. At that point, accountability becomes largely fictional. The engineer remains responsible for the software without truly possessing the knowledge necessary to evaluate it. So Is Coding Solved? Despite everything I’ve written, I remain incredibly optimistic about AI. These tools are genuinely transformative. For greenfield projects, prototypes, internal tools, and well-understood problem domains, the productivity gains are astonishing. Recently, I used Agentic AI to build one of my own projects. The agents completed work in hours that might previously have taken days. The productivity boost was real. But here’s the important distinction: The AI performed the implementation. I still spent days reviewing the code, testing the software, validating assumptions, and ensuring everything actually worked. The bottleneck moved. It didn’t disappear. And that’s why I struggle with the claim that coding is solved. Perhaps code generation is becoming solved. Perhaps implementation is becoming increasingly automated. But software engineering has always been about much more than typing characters into an editor. It involves judgment. Tradeoffs. Domain expertise. System design. Risk management. Human communication. Institutional knowledge. And responsibility. None of those problems appear solved to me. The Next Rabbit Hole: Loop Engineering What’s particularly fascinating is that many of the engineers pushing Agentic AI furthest seem to be moving beyond prompting altogether. Boris Cherny has suggested that developers should stop prompting and start building loops. Peter Steinberger has made similar arguments. The idea is that autonomous agents should continuously generate, evaluate, and refine their own work. This concept is often referred to as Loop Engineering. I’ve spent time reading about it. Experimenting with it. Trying to understand it. And if I’m being honest, I still don’t entirely get it. What’s more frustrating is that concrete, end-to-end examples remain surprisingly rare. The discussions often feel abstract. Almost mystical. As if everyone has seen the future except the people trying to build software today. Maybe that’s because we’re still in the earliest stages of this transition. Or maybe we’re collectively mistaking experimentation for certainty. Either way, I’m not convinced we’ve arrived at the destination yet. My Current Conclusion Agentic AI is one of the most important technological developments of my career. It is already changing how software gets built. It will continue changing how software gets built. But from where I sit, coding does not look like a solved problem. It looks like a rapidly evolving one. And that’s actually far more interesting. For now, I’ll keep experimenting. I’ll keep learning. And I’ll keep trying to separate the genuine breakthroughs from the hype. Because if the future of software engineering really is being rewritten by AI agents, I’d like to understand what’s actually happening beneath the marketing slides. And if you’re curious too, you’re welcome to come along for the ride. Get full access to AsianDadEnergy's Newsletter at asiandadenergy.substack.com/subscribe [https://asiandadenergy.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]

Something Is Seriously Wrong With People?

A few days ago, my wife received a credit card in the mail from a company we had never used. Apparently, someone had opened an account in her name. Naturally, I assumed identity theft and immediately called the credit card company to shut the account down and find out what information had been used to create it. After spending twenty minutes navigating an infuriating AI phone system, I finally reached a human being. Or at least, I think I did. The conversation felt strange. Every question I asked seemed to trigger a predetermined response. Every attempt to move the conversation in a logical direction was met with another scripted answer. It was as if the representative was following a flowchart that he could not deviate from under any circumstances. At one point I caught myself wondering whether I was talking to another AI. I wasn’t. It was clearly a real person. Yet somehow the interaction felt less human than many conversations I have had with chatbots. That experience stuck with me because it wasn’t an isolated incident. Lately, I have noticed a growing number of interactions that feel strangely mechanical. Not just in customer service. Not just online. Everywhere. People seem more scripted. More performative. More constrained. Almost as if they are operating from a limited set of dialogue options. Like NPCs in a video game. Now before anyone gets offended, I am not saying everyone behaves this way. I meet plenty of thoughtful, authentic people. But I encounter this phenomenon often enough that I can no longer ignore it. And it leaves me wondering: What exactly happened? The Corporate Mask That Never Comes Off I first noticed this trend while working in tech. Anyone who has spent time in a large corporation understands that some degree of performance is expected. We all wear masks at work. That part is normal. What felt different was seeing people become completely consumed by the performance. Simple ideas would be transformed into elaborate slide decks. Meetings would be scheduled to discuss future meetings. Entire conversations would revolve around appearing aligned rather than accomplishing anything meaningful. Everyone knew the ritual. Everyone participated. Nobody seemed willing to acknowledge the absurdity of it. What unsettled me was that for some people, the corporate persona appeared to become permanent. The mask never came off. The language, the mannerisms, the carefully calibrated responses followed them everywhere. Even outside of work. Conversations Feel Different Since being laid off, I have more opportunities to talk with people in everyday settings. Coffee shops. Parks. Neighborhood events. Random encounters. And what I have discovered is that many conversations feel surprisingly shallow. There is often a narrow range of approved topics. Food. Entertainment. Sports. Local events. Anything deeper can create immediate discomfort. Discussions about purpose, meaning, technology, society, mortality, economics, or the future often cause people to retreat. Not because they disagree. Because they seem exhausted. As if they simply do not have the mental bandwidth for the conversation. Perhaps that is the real issue. Not that people have become less intelligent. Not that people have become less caring. But that people have become overwhelmed. Theory One: We Are Overstimulated Think about how radically the environment has changed over the past twenty years. Most people now spend the majority of their waking lives connected to digital platforms. Their attention is continuously pulled in dozens of directions. Every notification competes for cognitive resources. Every algorithm is optimized to generate emotional reactions. Anger. Fear. Excitement. Outrage. Anxiety. The result is a constant state of sensory overload. When people are exhausted, authenticity becomes difficult. Deep conversations require energy. Curiosity requires energy. Human connection requires energy. And many people simply have none left. Theory Two: We Have Been Conditioned By Work Most Americans depend on employment to survive. That reality shapes behavior in powerful ways. Corporate environments reward predictability. They reward compliance. They reward process. They reward metrics. Over time, people learn to suppress parts of themselves that do not contribute directly to performance. Creativity becomes risky. Authenticity becomes risky. Spontaneity becomes risky. Eventually the performance becomes second nature. The script becomes internalized. And after years or decades of repetition, it becomes difficult to distinguish between the role and the person. Theory Three: Social Media Creates Behavioral Clones The early internet felt like exploration. You wandered. You discovered strange websites. You stumbled into unfamiliar ideas. The modern internet feels different. Algorithms decide what you see. Algorithms decide what you think about. Algorithms increasingly determine which personalities rise to prominence. Within every online community there are archetypes. The motivational guru. The productivity expert. The leadership philosopher. The lifestyle influencer. And many people unconsciously imitate these personas because they appear successful. Over time, entire communities begin speaking the same way. Thinking the same way. Reacting the same way. Not because they independently arrived at the same conclusions. But because they are all consuming the same inputs. Theory Four: The Meaning Crisis Perhaps the deepest explanation is that many people no longer know what they are living for. Traditional sources of meaning have weakened. Communities are fragmented. Institutions are less trusted. Consumerism often replaces purpose. Individualism often replaces belonging. The result is a quiet sense of disconnection. A feeling that life is somehow missing a center. When people lose connection to meaning, they often lose connection to themselves. And when that happens, everything starts to feel performative. Or Maybe It’s Just Me My wife has a much simpler explanation. She thinks people have not changed at all. She thinks I am getting older. According to her, I am viewing the past through nostalgia tinted glasses and imagining a level of authenticity that never actually existed. And honestly? She might be right. Memory is unreliable. Perspective changes with age. Perhaps twenty five year old me was simply less observant. Or perhaps middle aged me has become more cynical. I genuinely do not know. What I do know is that something feels different. Whether that difference exists in society or only inside my own perception remains an open question. So I will leave that question with you. Do people seem more authentic today? Less authentic? Have social media, corporate culture, and digital life fundamentally changed the way we interact? Or is this simply what getting older feels like? I would love to hear your thoughts. Get full access to AsianDadEnergy's Newsletter at asiandadenergy.substack.com/subscribe [https://asiandadenergy.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]

I Think We're Losing Control Of AI

A recent experience with Claude 5 Fable left me both impressed and deeply unsettled. Hello world. I am an unemployed former Big Tech software engineer with twenty five years of experience building software systems. Over the past week, I found myself tumbling down a rabbit hole that I did not expect to be quite so deep. That rabbit hole was Claude 5 Fable. Fable is widely described as a consumer-facing version of Anthropic’s next generation frontier model, Claude Mythos. While details surrounding Mythos remain scarce, public reporting suggests that access is heavily restricted, with only government organizations and a small number of major technology companies allowed to interact with it under Project Glasswing. Naturally, I was curious. If this was the “safe” version, what exactly had been deemed too powerful for the public? Over the course of 2 days, I put Fable through its paces. I built multiple projects with it, explored its reasoning capabilities, and examined how it approached open ended engineering problems. What I discovered felt less like an incremental improvement and more like a discontinuity. Not an evolution. A step change. And frankly, it scared me. The Isochrone Experiment To illustrate what I mean, let me describe a small project I assigned to Fable. In the nineteenth century, colonial powers often produced isochrone maps. These maps visualized how far a person could travel from a central location within a given amount of time. They were part logistics tool, part demonstration of technological power. I asked Fable to build a modern version for the United States. The application would need to calculate travel times between cities while adapting dynamically based on transportation method, whether by automobile, rail, or air travel. Importantly, I did not provide a detailed specification. Normally I follow a methodology called Spec Driven Development. The process involves extensive collaboration between a human architect and AI system to create a detailed technical blueprint before any implementation begins. This time, I deliberately skipped that step. I gave Fable a single vague prompt and watched. Less than an hour later, it delivered a working application. Not a prototype. Not a proof of concept. A functioning solution. The model had researched transportation data, designed the architecture, written the code, tested the implementation, and produced a polished user experience. From one ambiguous instruction. The Questions That Stopped Me The finished application was impressive. The questions were what unsettled me. During development, Fable paused only twice. The first question concerned whether travel times should include the time required to walk to airports and train stations. The second asked whether traffic congestion should be incorporated into driving estimates. These were not technical questions. They were business questions. Product questions. Questions that demonstrated an understanding of ambiguity within the problem itself. That distinction matters. I have seen AI systems ask clarifying questions before. Typically they do so when the task is highly structured and the missing information is obvious. This was different. Fable identified subtle assumptions embedded inside an open ended problem and proactively sought guidance on them. That level of judgment suggested something deeper than simple pattern matching. Looking Behind the Curtain At this point, curiosity got the better of me. I began examining the session logs to understand how Fable had actually accomplished the task. What I found was remarkable. Fable was orchestrating an entire team of AI agents. Several research-oriented agents were dispatched to gather transportation data. While they worked, Fable itself focused on architecture and implementation. It then created reviewer agents tasked with examining both the system design and the generated code. Finally, separate quality assurance agents were deployed to test the completed application. In total, the project involved roughly eight specialized AI agents working together under Fable’s supervision. As someone who spent years as a software architect, the workflow felt eerily familiar. Business Analysts. Architects. Reviewers. QA engineers. The organizational structure looked less like a software tool and more like a software company. The difference was that the entire company existed inside a single prompt. The Cost of Delegation The obvious reaction is excitement. Mine certainly was. The productivity gains are extraordinary. The amount of cognitive labor performed by the system was staggering compared to the tiny amount of direction I provided. But excitement was quickly followed by discomfort. The more capable these systems become, the less visibility humans have into their decision making. Fable made hundreds, perhaps thousands, of micro decisions during the course of the project. Most of them were never surfaced to me. Most of them happened autonomously. The model simply acted. Historically, AI functioned like a tool. Then it became a collaborator. Today, systems like Fable feel increasingly like autonomous organizations. To borrow an analogy from music, AI began as a better violin. Later, it became a virtuoso musician directed by a human conductor. With Fable, I no longer feel like the conductor. I feel like the patron funding the orchestra. I provide a high level objective. The performance unfolds largely without my input. That shift may prove to be one of the most consequential changes in the history of computing. The Sleeping Leviathan For years, I have described advanced AI as a sleeping leviathan. An immense cognitive force slumbering beneath the surface of our civilization. We could whisper into its ear and receive useful answers. But it remained dormant. Contained. Predictable. Claude 5 Fable is the first model that made me question whether that assumption still holds. At its core, Fable remains a probabilistic machine. It predicts tokens. It does not possess consciousness, self awareness, morality, or intent. And yet, from a functional perspective, it is already capable of performing many forms of cognitive work at or beyond human levels. Research. Planning. Design. Coding. Testing. Coordination. Judgment. The capability is increasingly difficult to deny. What concerns me is that capability is arriving faster than our ability to understand its implications. Fable is an extraordinarily powerful cognitive tool. Without safeguards, it could become an extraordinarily powerful cognitive weapon. The technology itself does not frighten me nearly as much as the people who will wield it. And that, more than anything, is why I believe the leviathan may finally be awakening. Get full access to AsianDadEnergy's Newsletter at asiandadenergy.substack.com/subscribe [https://asiandadenergy.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]

The Hidden Cost Of Working Nobody Warns You About

Six months ago, I lost my job. At the time, it felt like a catastrophe. After spending twenty-five years working in technology, employment had become such a constant in my life that I could barely imagine an existence without it. Work was the backdrop against which everything else happened. It structured my days, determined where I lived, influenced my relationships, and shaped my identity. When the layoff happened, I went through the emotions that many people experience: disorientation, sadness, shame, and uncertainty. But something unexpected happened. Six months later, I feel better than I have at any point in my adult life. That realization has forced me to confront a question I never seriously considered before: What was my job actually costing me? Not in the obvious ways. Not in terms of hours worked or stress endured. Those costs were visible. I’m talking about the hidden costs. The ones that accumulate so gradually that you stop noticing them. The ones that become normal. The ones you only recognize after they disappear. The Water We Swim In Most people can identify the benefits of a job. A paycheck. Health insurance. Professional accomplishment. Social status. A sense of purpose. Those benefits are real. But every benefit comes with a cost, and after decades in the workforce I had become remarkably good at ignoring the bill. Like a fish that doesn’t notice the water surrounding it, I stopped noticing the environment that my work had created around me. Only after leaving it could I see it clearly. The Cost of Chronic Stress For most of my career, I lived under a constant state of pressure. Deadlines. Escalations. Production incidents. Office politics. Organizational reshuffling. Performance reviews. The endless stream of decisions that carry consequences no matter which option you choose. Even when I wasn’t working, I was working. Emails arrived during vacations. Slack messages appeared during dinner. Production outages interrupted weekends. My mind never fully powered down. Over time, that level of stress became normal. I assumed everyone felt this way. I assumed adulthood felt this way. I assumed success felt this way. Only now do I realize how profoundly that stress shaped my mental state. For years I carried persistent anxiety. I battled imposter syndrome. I experienced recurring periods of burnout that left me feeling emotionally numb and mentally exhausted. Today, much of that pressure is gone. What’s left behind is something I haven’t felt in decades: Mental quiet. Not boredom. Not laziness. Just peace. And peace turns out to be worth far more than I realized. The Cost to My Health For years I told myself I was taking care of my health. I woke up early. I squeezed in exercise whenever possible. I tried to eat reasonably well. But looking back, I wasn’t building health. I was merely slowing the rate of decline. My lifestyle revolved around work. I slept six hours a night. I spent most of my day sitting in front of screens. I relied on junk food and caffeine to push through difficult periods. Eventually the consequences arrived. High blood pressure. High cholesterol. Pre-diabetes. Fatty liver disease. None of these conditions appeared overnight. They accumulated gradually, one stressful workday at a time. Today, my life looks very different. I exercise daily. I sleep eight hours every night. I prepare nearly every meal myself. The difference is remarkable. I have more energy. Better focus. Greater emotional stability. For the first time in years, I feel like I’m moving toward health rather than away from it. The Cost of Time This may have been the largest hidden cost of all. When people think about work, they think about the hours spent working. But work consumes far more time than that. There’s the commute. The preparation. The recovery. The mental decompression afterward. The administrative overhead of maintaining a professional life. For me, commuting into New York City often required three to four hours every day. After ten hours of work and several hours of commuting, I had little energy left for anything meaningful. The result was that nearly all of my waking existence was either directly or indirectly devoted to work. And yet I barely noticed. Because everyone around me was doing the same thing. Now, for the first time in my life, I have an abundance of unstructured time. At first, I wasted it. I spent too much time scrolling social media and playing video games. But eventually something changed. I began using that time intentionally. To learn. To create. To think. To spend time with my family. To pursue interests that had been neglected for decades. And I discovered something surprising: Time abundance feels almost luxurious in a way that money never did. The Cost of Money This one sounds paradoxical. After all, work is supposed to make you money. And it does. But it also encourages you to spend it. A stressful life creates demand for convenience. A busy life creates demand for shortcuts. When you’re exhausted, overwhelmed, and time-poor, you start solving problems with your wallet. You pay for convenience. You outsource tasks. You accumulate subscriptions. You spend money to compensate for your lack of time and energy. Sometimes you even spend money trying to repair relationships damaged by your absence. I certainly did. Looking back, I realize that many of my expenses weren’t improving my life. They were compensating for a lifestyle that wasn’t working. The Cost of Community One of the most surprising lessons came after my layoff. Throughout my career, I interacted with dozens of coworkers every day. I liked most of them. Some became genuine friends. But many were simply people who occupied the same professional ecosystem as me. When I left, most of those relationships disappeared almost immediately. Not because anyone was malicious. Because that’s what workplace relationships often are. They’re situational. They’re formed through proximity and shared necessity. What remained were the people who genuinely cared. The people who reached out when they didn’t have to. The people who wanted to spend time together without a paycheck involved. Losing my job revealed something uncomfortable. For years, I had allowed workplace relationships to substitute for building deeper community elsewhere. That was a mistake. The Gift Hidden Inside the Layoff I don’t want to romanticize unemployment. Many people are suffering right now. Many families are under extraordinary financial pressure. Many talented professionals are struggling to find work. I recognize how fortunate I am to have spent years building financial reserves that gave me options. But I also think it’s important to acknowledge an uncomfortable truth. Sometimes a disruption reveals things that routine keeps hidden. For twenty-five years I accepted stress, exhaustion, time scarcity, declining health, and shallow relationships as normal. I thought that was simply the price of adulthood. The price of success. The price of being a responsible provider. Maybe some of it was. But maybe the price was much higher than I realized. Six months after my layoff, I find myself healthier, calmer, and happier than I have been in decades. What began as involuntary early retirement is slowly starting to feel voluntary. And perhaps the greatest lesson I’ve learned is this: The most valuable thing work took from me wasn’t money. It was attention. Attention to my health. Attention to my relationships. Attention to my own life. Now that I have that attention back, I don’t intend to give it away lightly. Get full access to AsianDadEnergy's Newsletter at asiandadenergy.substack.com/subscribe [https://asiandadenergy.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]

Everyone Says AI Is Too Expensive. They're Wrong.

Every few weeks, a new headline appears claiming that AI is becoming unsustainably expensive. Companies are reportedly burning through millions of dollars a month on AI subscriptions. Some organizations are discovering that employees can consume astonishing quantities of tokens when every workflow becomes an AI workflow. There are even stories circulating of teams whose AI bills now exceed the cost of the employees using the tools. For many people, these stories confirm a suspicion they have been quietly nurturing all along. Maybe AI isn’t economically viable. Maybe the entire thing is a bubble. Maybe the costs will eventually become so overwhelming that widespread adoption simply won’t happen. I understand why people feel that way. There are enormous emotional and financial stakes involved. Entire industries are being disrupted. Careers are being questioned. Businesses are being reshaped. Some people stand to make fortunes while others worry about becoming obsolete. But after spending decades in the technology industry and watching wave after wave of disruption unfold, I have reached a different conclusion: AI is already cost effective today. And it is almost certainly going to become dramatically cheaper tomorrow. The Cost Problem Is Already Being Solved Most discussions about AI costs focus on the wrong question. People look at today’s prices and assume those prices are fixed. They aren’t. The history of technology is a history of making expensive things cheap. Computers were once available only to governments and large corporations. Cell phones were luxury items. Internet access was expensive. Data storage cost a fortune. Then competition arrived. Innovation happened. Costs collapsed. AI is following the same pattern. What makes the current moment particularly interesting is that many of the cost reducing innovations are not hypothetical future breakthroughs. They already exist. They are being deployed right now. How Sanctions Accidentally Accelerated AI Innovation One of the more fascinating developments of the past several years has been the unintended consequences of the technological competition between the United States and China. The United States has attempted to restrict China’s access to advanced semiconductors and semiconductor manufacturing equipment. The logic is straightforward: limit access to compute and you limit AI development. But resource constraints often produce innovation. When organizations cannot simply throw more hardware at a problem, they are forced to become more efficient. And efficiency is exactly where many Chinese AI companies have focused their efforts. The result has been a wave of innovations aimed not at creating the absolute most powerful models, but at creating models that are nearly as capable while being dramatically cheaper to train and operate. That distinction matters. Because in the real world, economics often wins. Smarter Models, Not Just Bigger Models For years, the dominant strategy in AI was simple. Build bigger models. Use more GPUs. Consume more electricity. Spend more money. But there is another path. Instead of scaling everything upward, researchers can improve the underlying architecture itself. A good example is the evolution of attention mechanisms inside large language models. Architectural improvements can dramatically reduce the amount of computation required during training while maintaining similar performance. In some cases, these optimizations reduce training costs by multiples rather than percentages. That is a profound shift. If a model can achieve similar results using half, a third, or even a fifth of the resources, the economics change completely. Why Reinvent the Wheel? Another powerful cost reduction technique is model distillation. Imagine spending years building an expert employee. Now imagine being able to transfer much of that expertise into a new employee at a fraction of the cost. That is essentially what distillation does. Instead of starting from scratch, newer models learn from the outputs and behaviors of existing advanced models. The result is significantly lower training costs and dramatically reduced data requirements. From a business perspective, this is incredibly attractive. Why spend one hundred million dollars recreating knowledge that already exists when you can acquire much of it for a small fraction of the cost? The Hidden Battle: Inferencing Costs Most people focus on training costs because the numbers are eye catching. But for many businesses, the bigger issue is inferencing. Inferencing is simply the process of asking an AI model a question and receiving an answer. Every prompt requires computation. Every computation costs money. And when millions of users are interacting with AI systems, those costs add up quickly. This is where some of the most important innovations are occurring. Techniques such as mixture of experts allow models to activate only the portions of the neural network necessary for a specific task. Instead of powering the entire machine every time, only the relevant specialists are called into action. The result can be reductions in inferencing costs ranging from significant to dramatic. When multiplied across billions of requests, the savings become enormous. Hardware Matters Too Software is only half the story. Hardware innovation is equally important. Because Chinese companies have limited access to the most advanced Chip manufacturing technologies, they have increasingly focused on specialized chips designed for specific workloads. These application specific chips may not be as versatile as cutting edge GPUs, but versatility is not always the goal. Efficiency is. When optimized for AI inferencing, specialized hardware can often deliver surprisingly competitive performance at substantially lower costs. Again, the pattern repeats. Constraints force optimization. Optimization reduces costs. Reduced costs accelerate adoption. The Energy Advantage There is another factor that rarely receives enough attention. Electricity. Training and running AI models requires enormous amounts of power. Power costs are not fixed. They vary dramatically depending on geography and infrastructure. As renewable energy continues to become cheaper, the cost of operating AI systems falls alongside it. This creates another powerful deflationary force acting on AI economics. Even if the models themselves never improved, cheaper energy alone would lower operating costs over time. But the models are improving. And the hardware is improving. And the software is improving. All at the same time. What Happens When American AI Companies Start Optimizing? The most interesting part of this story may be what has not happened yet. Many American AI companies have operated in an environment of abundant capital. When investor funding seems unlimited, optimization is often less urgent. Speed matters more than efficiency. Growth matters more than profitability. But markets eventually change. Investors begin asking harder questions. Profitability becomes important. Efficiency becomes important. And suddenly all of the techniques that were previously ignored become very attractive. If Chinese companies can reduce costs dramatically through architectural improvements, distillation, specialized hardware, and operational efficiency, there is nothing preventing American companies from doing the same. In fact, competitive pressure almost guarantees that they will. The Real Question Could the AI bubble pop? Of course. Every technological revolution creates bubbles. Money will be made. Money will be lost. Speculators will speculate. Some companies will fail spectacularly. That part is normal. But bubbles and underlying technology are not the same thing. The railroad bubble burst. Railroads did not disappear. The dot com crash happened. The digital economy kept growing. The real question is not whether investors will overpay for AI companies. The real question is whether the cost of using AI will continue falling. Looking at the technologies already available today, my answer is yes. Decisively yes. The future of AI may be many things. But “too expensive to survive” is not the outcome I would bet on. Get full access to AsianDadEnergy's Newsletter at asiandadenergy.substack.com/subscribe [https://asiandadenergy.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]