Why I Stopped Teaching My Kids How To Code - And Why You Should, Too!

Why the future will be a future of energy wars, why software engineering as a discipline will disappear (and why the common wisdom is wrong)

This newsletter is brought to you by “The 7 Data Product Strategies” - the ebook for product managers and entrepreneurs who want to build amazing data-heavy products.

Recent praise for the book: "I love how Sven is looking at data and data strategy from a different angle - as a business opportunity. He gives some great tips on how to find these valuable datasets within your company and then how to take advantage of them. But these proposals are not technical in nature, he analyses the problem from the business model perspective.” - Jacek Majchrzak, Lead Architect, Facilitator & Data Wizard

---

My kids are super interested in programming; of course, you can program games, after all. And Minecraft mods… So, I started to teach them how to use Scratch, an amazing programming education initiative from MIT.

People say it’s hard to predict AI progress over the next years due to exponential drivers and the uncertainty of not knowing. And I’m 100% with that. It’s hard to tell my kids what they should learn and should invest in because the future of tech seems too hard to grasp.

But lately I’ve settled on one truth I can tell my kids: Their future won’t involve programming. It won’t involve software development or data engineering. In fact, I stopped teaching my kids how to code, and I think you should, too!

The world our kids will grow up in won’t need coders, programmers, or software engineers anymore. And today, I want to tell you why most people are mistaken about their views of the future of these disciplines. And what, instead, will the drivers of the future be: brute force and energy consumption?

Gold Mining

I recently learned about gold mining. It is an industry largely hidden from the forefront of newspapers, an industry where we’d probably not send out kids to work.

In the realm of gold mining, a revolution has quietly unfolded. Gone are the days of prospectors panning for gold in rushing streams. Today, massive machines process tons of earth, extracting tiny amounts of gold through sheer brute force. This shift from human labor to energy-intensive processes marks a fundamental change in how we approach resource extraction.

"The future is already here – it's just not evenly distributed," said William Gibson, the prescient science fiction author. His words ring true as we stand on the precipice of a similar revolution in software development. Just as gold mining evolved, we're witnessing a transformation in how we create digital products.

The parallels are striking. In software, we're approaching a point where brute computational power is poised to replace sophisticated human work. Artificial Intelligence (AI) now allows us to "throw" vast amounts of compute and data at problems, bypassing the need for human intricacy and craft in coding. This isn't just a minor shift in how we write software—it's a fundamental revolution in how we build digital products.

This idea aligns perfectly with Sam Altman's famous prediction of billion-dollar startups run by a single person. In this new landscape, AI, fueled by sheer computational energy, will enable individuals to build what once required entire teams - by replacing vast armies of software engineers with brute force energy.

But objections to a future without software developers are strong, so let’s consider them one by one.

---

Part 1: Common Objections to AI Replacing Software Engineering

There are tons of objections on why AI will never replace software engineers, why there still will be a need for higher-skilled software engineers, and so on.

However, I’ve come to notice that those arguments are always based on two fallacies.

The Value Fallacy: People love to argue that software engineering is essential and will simply evolve as a discipline. But this central idea misses the central picture: software engineering is just a tool. It’s a tool to create products and software, and that’s it. AI is fundamentally changing how we build products, not just how we write code. At its core, the shift is from human expertise to brute computational power.

The AI Won’t Ever Fallacy: People love to argue based on the current limitations of AI. They implicitly make three assumptions: 1. AI is gonna stay close to as expensive as it is today. 2. AI is gonna be limited by the same constraints as it is today 3. Future AIs will look like LLMs. The thing is, none of this is true. AI model costs are falling exponentially 2. The AIs we have today are the worst ones you will ever see 3. LLMs are not the end game!

Let’s see how specific objections play out.

1. "AI is bad at solving complex software problems."

At first glance, it might seem that AI struggles with complex, creative, or ambiguous software problems. This objection echoes the skepticism that once surrounded calculators in mathematics. Just as some initially doubted that machines could outperform human mental arithmetic, many today question AI's ability to tackle intricate software challenges.

However, this perspective overlooks a crucial reality. As Demis Hassabis, CEO of DeepMind, aptly puts it, AI can solve complex problems that humans can't, not because it's smarter than humans, but because it can consider vastly more possibilities. This insight cuts to the heart of AI's true advantage.

The real issue isn't that AI struggles with complex problems; it's that developers themselves often grapple with these challenges. AI is starting to excel in areas where humans falter, using its immense computational power to simulate and analyze possibilities at a scale no human could match. To visualize this, imagine a "Problem Complexity vs. Solution Space" diagram. While a human developer might explore a limited area of potential solutions, AI can sweep through an exponentially larger space, uncovering innovative approaches we might never consider.

A prime example of this is Google's use of AI in chip design. In this complex, creative field, AI outperformed human experts by exploring unconventional layouts and optimizations. The AI-designed chips were not only superior in performance but were completed in a fraction of the time it would take a human team.

But you can make this true today by using AI to tackle any problem in which you don’t have expertise. AI is, of course, gonna be 10x better than you. And why shouldn’t it? So, of course, AI will be able to be 10x more knowledgeable than 99% of software engineers today.

AI's advantage lies in its ability to apply brute-force computing to handle complexities that would take humans much longer if they could manage them at all. With falling model costs, solving complex problems becomes a function of energy - sheer brute force like gold mining.

2. "AI can't handle ambiguity and vague requirements."

A common objection to AI in software development is that it struggles with ambiguous or evolving project requirements. This concern stems from the belief that AI, being a machine, needs precise instructions to function effectively. However, this perspective overlooks both the nature of modern AI and the realities of software development.

First, let's consider the shifting sands of software development. Project requirements have always been a challenge, even for human teams. The concept of "requirements volatility" in software engineering acknowledges that change is constant. Projects evolve, stakeholders change their minds, and market conditions shift. Human developers have long grappled with these moving targets.

Andrej Karpathy, former AI Director at Tesla, offers a crucial insight: The first step in tackling any problem is to collect, clean, and organize your data. AI thrives on data, even when it's messy or ambiguous. This statement cuts to the heart of AI's strength in dealing with uncertainty.

While it's true that AI isn't perfect at interpreting vague inputs today, its ability to process and model enormous data sets allows it to infer product requirements with increasing accuracy. AI systems can analyze vast amounts of user feedback, market trends, and historical project data to make informed predictions about what a product needs, even when human stakeholders are uncertain.

The thing is, software engineers aren’t great at understanding vague requirements either! They also need tons of context, product management, and the like to get to the core of the requirement. In essence, software engineers need to do what AI is doing really great: collecting lots of data and making sense of it.

3. "AI lacks domain-specific knowledge and can't replace human expertise."

A prevalent concern in the AI debate is that these systems lack the deep, nuanced, industry-specific knowledge that human experts bring to the table. This objection often stems from the belief that human expertise, honed over years of experience, is irreplaceable. However, this perspective may be rooted more in our reverence for human cognition than in the realities of AI capabilities.

Historically, we've seen machines surpass human experts in domains once thought to be the exclusive purview of human intelligence. Chess grandmasters, Go champions, and even medical specialists have found themselves outperformed by AI in specific tasks. These examples challenge our assumptions about the irreplaceability of human expertise.

Moreover, human experts can suffer from what psychologists call "expert blindness" – a tendency to overlook novel solutions or perspectives due to entrenched ways of thinking. AI, free from such cognitive biases, can approach problems with a fresh perspective.

As Fei-Fei Li, Co-Director of Stanford's Human-Centered AI Institute, insightfully notes: AI doesn't need to replicate human intelligence. It just needs to solve problems and provide insights, possibly using very different cognitive strategies."This statement encapsulates a crucial point: AI's strength lies not in mimicking human expertise but in leveraging its unique capabilities to handle domain knowledge differently – and often more effectively.

A prime example of this is in the field of medical diagnosis. AI systems have demonstrated the ability to outperform human experts in diagnosing certain conditions, particularly in image-based specialties like radiology. What's more, AI can integrate knowledge from multiple medical specialties simultaneously, something that would be challenging even for the most experienced human doctor.

AI may lack domain-specific expert knowledge for now, but it won’t forever. And the thing is, even today, it doesn’t need to. AI can work differently and still get a better or similar result. If you get the same result as an expert software developer with the click of a button, why would you hire an army of expensive software developers in the first place?

4. "AI doesn't understand the big picture and can't make long-term decisions."

A common objection to AI in software development is the belief that it cannot grasp the big picture or make the long-term decisions necessary for effective software architecture and system sustainability. This view stems from the traditional role of software architects, who are valued for their ability to make significant design decisions that shape a system's future.

However, this perspective may overestimate the long-term effectiveness of human decision-making while underestimating the potential of AI in this domain. Human architects, despite their expertise, are not immune to mistakes. The concept of "architectural debt" – where initial design decisions lead to compounding issues over time – is a testament to the limitations of human foresight in system design.

Grady Booch, Chief Scientist for Software Engineering at IBM Research, provides a useful framework: All architecture is design, but not all design is architecture. Architecture represents the significant design decisions that shape a system, where significance is measured by the cost of change. This definition highlights that the essence of good architecture lies not in human intuition but in making decisions that minimize the cost of future changes.

This is where AI's true advantage comes into play. While AI may not "understand" the big picture in the way a human does, it excels at using computational brute force to test and refine architectures dynamically in real time. AI can simulate thousands of potential architectural decisions, evaluate their long-term implications, and optimize for factors like scalability, maintainability, and performance – all at a speed and scale that no human architect could match. Consider an "Optimization Space" graph. While a human architect might explore a limited number of architectural solutions based on experience and intuition, an AI system can sweep through an exponentially larger solution space, uncovering innovative approaches that humans might never consider.

---

Part 2: The Implications of AI Replacing Software Engineering

The true battle of the future won't be about who has the best developers or the most sophisticated code—it will be about who controls the most energy. Progress will be a function of energy production because brute-force computational power will drive innovation and product development.

One key fallacy is to think that software engineering will be “upgraded.” Think about this for a second: when was the last time a gold mining expert made the front page of any news outlet? Gold mining needs experts, too! Gold mining as a discipline has been upgraded. But it has been upgraded to a place where 99% of the work is done by brute force and thus requires very limited human capital at all.

Software engineering as a discipline will be upgraded into the shadows. In the future it’s gonna be easy to create great digital products just based on ideas, sketches and data.

AI: The New Electricity of Innovation

In the realm of technological revolutions, few statements have captured the imagination quite like Andrew Ng's bold declaration: AI is the new electricity. Ng, a prominent figure in the AI world and co-founder of Google Brain, wasn't merely making a catchy comparison. He was drawing a profound parallel between two transformative forces that have reshaped – and are reshaping – the very fabric of our society and economy.

To truly appreciate the weight of Ng's metaphor, we must first decode its meaning in the context of software development and product creation. Just as electricity revolutionized every industry it touched during the Industrial Revolution, AI is poised to fundamentally alter how we approach innovation, problem-solving, and product development in the digital age.

The parallels between electricity and AI are striking. In the late 19th and early 20th centuries, electricity transformed manufacturing, communication, and daily life. Factories that once relied on steam power and complex systems of belts and pulleys could now use electric motors, dramatically increasing efficiency and flexibility. Similarly, AI is revolutionizing software development and product creation.

Tasks that once required teams of programmers can now be accomplished by AI systems, allowing for rapid prototyping, testing, and iteration. Consider how electricity became an invisible, essential utility – so seamlessly integrated into our lives that we only notice it when it's absent. AI is following a remarkably similar trajectory. From the algorithms that power our search engines to the recommendation systems that curate our entertainment, AI is becoming an invisible force that permeates nearly every digital interaction we have. The democratization of innovation is another crucial parallel. Electricity enabled small-scale entrepreneurs to compete with large industries by providing access to powerful, affordable tools. Today, AI is playing a similar role in the tech landscape. Cloud-based AI services and open-source machine learning libraries are allowing individual developers and small startups to create sophisticated, AI-driven products that can compete with those of tech giants. Just as the electrical revolution required a new infrastructure of power plants and distribution grids, the AI revolution is built on a foundation of data centers and cloud computing networks. Access to these resources is becoming as crucial for innovation in our era as access to electrical power was a century ago. Companies and countries that control these resources may well become the powerhouses of the AI age, much like those who controlled electrical infrastructure shaped the 20th-century economy.

1. For Individuals: A Radical Shift in Skillsets

The tech industry has always been characterized by rapid change, but the rise of AI in software development heralds a transformation more profound than any we've seen before. We're witnessing nothing less than the twilight of traditional programming as we know it.

Historically, programming has evolved from low-level machine code to high-level languages, each step abstracting away complexity. But AI represents a quantum leap in this progression. As Andrej Karpathy, former Director of AI at Tesla, notes, Software 2.0 is writing software that writes software. This succinct statement encapsulates the seismic shift occurring in our field.

With AI increasingly capable of handling complex coding tasks, from bug fixing to entire system design, the role of the human developer is being radically redefined. We're moving from an era of programmers to one of AI-augmented developers. This doesn't mean the end of human involvement in software creation but rather a shift in focus. As Grady Booch, Chief Scientist at IBM Research, puts it, The future of coding is no coding at all.

In this new landscape, the most valuable skills revolve around managing and directing AI systems rather than writing code line by line. The ability to effectively prompt AI and understand its capabilities and limitations becomes crucial. It's akin to learning a new language – the language of AI collaboration.

It is hard to say what the individual implications will be, but what I can say is that you can safely stop learning how to code.

2. For Companies: Restructuring Product Development

The integration of AI into software development isn't just changing how we code; it's fundamentally reshaping how companies structure their product development processes. We're witnessing the dawn of a new era in which AI-driven development teams are becoming the norm, not the exception.

Traditionally, software development teams were structured around specialized roles: front-end developers, back-end developers, QA testers, and so on. However, as Satya Nadella, CEO of Microsoft, notes, AI is not just the next big wave in computing — it's the next major turning point in human history. This turning point is manifesting in how companies are restructuring their development teams.

With AI taking over many of the technical aspects of coding, companies are finding they can operate with much leaner teams. These AI-augmented teams are not just smaller; they're faster and more efficient. Software is eating the world, but AI is going to eat software."This process is accelerating the innovation cycle to unprecedented speeds.

Perhaps the most radical change is the rise of the AI-empowered solopreneur. Sam Altman's prediction of billion-dollar startups run by single individuals is becoming increasingly plausible. AI's ability to handle complex tasks allows individuals to create and manage products that would have required entire teams in the past.

However, this new paradigm requires companies to adapt their structures and processes. Hierarchies are flattening as AI takes over many middle-management functions. Decision-making is becoming more data-driven and decentralized. Looking ahead, we might envision a tech company where small, cross-functional teams of AI-augmented developers work alongside AI systems to rapidly prototype, test, and iterate on products. The bulk of the 'workforce' might be AI models, with humans acting as strategic directors and creative problem-solvers.

I know this is repetitive. But maybe it helps to drive home the point ;)

3. For Nations: Geopolitical and Economic Shifts

As we stand on the brink of an AI-driven future, nations around the world are waking up to a stark reality: the next great power struggle will be fought not over oil or rare earth minerals, but over the energy needed to fuel the AI revolution. This shift is set to redefine global politics and economics in ways we're only beginning to understand.

The parallels with the oil wars of the 20th century are striking. Just as control over oil reserves shaped global power dynamics for decades, access to abundant, cheap energy is poised to become the key determinant of national competitiveness in the AI era. As Vaclav Smil, renowned energy scientist, puts it, "The history of civilization is a history of energy transitions." We are now witnessing the dawn of perhaps the most significant energy transition yet.

In this new landscape, traditional measures of national power - GDP, military might, or even technological prowess - may take a backseat to a nation's capacity to produce and distribute energy. Countries blessed with abundant energy resources, or those able to innovate in energy production, stand to gain a significant edge. In this context, the nations that control the "electricity" will control the future. Consider the case of China. Already a leader in AI research, China is also the world's largest producer of renewable energy. This combination could prove potent, potentially allowing China to leapfrog other nations in AI capabilities. As Ryan Hass, a China expert, observes, China's dual strengths in AI and renewable energy could reshape global power dynamics in the coming decades. This shift is likely to exacerbate global inequalities, creating a widening AI divide. Nations with limited energy resources or underdeveloped infrastructure may find themselves left behind in the AI race. The challenge for these countries will be immense: how can they compete in a world where AI-driven productivity becomes the primary engine of economic growth?

Brute Force, Not Sophistication, Will Shape the Future of Software Development

As we stand at the precipice of this AI-driven future, let's return to our gold mining analogy. The shift we're witnessing in software development isn't just about automation—it's a fundamental replacement of human sophistication with brute computational force, much like how we transitioned from manual gold panning to energy-intensive industrial extraction.

Sam Altman's vision of billion-dollar startups run by single individuals is rapidly becoming a reality. In this new paradigm, energy truly becomes the currency of innovation. The pioneers of this new digital frontier won't necessarily be the most brilliant coders or the largest tech conglomerates. Instead, they'll be those who can most effectively harness and direct AI's brute force capabilities.

Imagine a day in 2030: A solo entrepreneur wakes up, gives a series of prompts to her AI assistant, and by lunchtime has prototyped, tested, and launched a new app used by millions. This isn't science fiction—it's the logical conclusion of the trends we're seeing today.

So, what do I teach my kids today? I don’t know! AI is hard to predict. I can tell you only that I don’t teach them how to code, and then I teach them everything about AI I have to give, and you should, too. Your kids, your employees, your customers.