There's a seductive narrative floating around right now: AI will let anyone do anything. Can't code? AI will write it for you. Don't understand design? AI will generate it. Never written a business plan? AI will draft one in seconds. It sounds liberating, but it hides a trap. AI doesn't help you start from zero. It helps you go faster when you already know where you're going.

The 70% illusion

Addy Osmani, an engineering leader at Google, has written extensively about what he calls the "70% problem" in AI-assisted coding. AI can rapidly produce about 70% of a solution, but the remaining 30%, which includes edge cases, security, production readiness, and maintainability, still requires genuine engineering knowledge. This pattern shows up everywhere, not just in software. AI can generate a marketing strategy, but knowing which channels actually work for your audience requires experience. AI can summarize a research paper, but evaluating whether the methodology is sound requires domain expertise. AI can draft a financial model, but spotting the assumptions that will blow up requires years of pattern recognition. The 70% looks impressive. It's the 30% that separates something useful from something dangerous.

The amplifier effect

Here's what the research consistently shows: AI tools help experienced people far more than beginners. Osmani puts it plainly: "Expertise is the foundation. AI tools are the amplifier. The stronger the foundation, the louder the amplifier becomes." When you bring genuine skill, judgment, and a systems-level perspective, AI becomes a force multiplier. An experienced developer can prompt AI in ways that yield senior-level results because they know what good output looks like. They can spot when AI is subtly wrong. They can take the 70% and finish the job. But if you don't have that foundation, you're amplifying nothing. Or worse, you're amplifying mistakes you can't even recognize.

What happens when you skip the fundamentals

A recurring thread on developer forums tells the same story: junior engineers who rely heavily on AI can produce code that appears to work but is built on deprecated APIs, poor patterns, or logic they don't understand. When it breaks, they're stuck. They can't debug what they never understood in the first place. This isn't just anecdotal. A 2025 MIT Media Lab study found that excessive reliance on AI-driven solutions may contribute to "cognitive atrophy," a reduction in critical thinking abilities. The study monitored neural activity in participants writing essays with and without AI assistance. Those who used ChatGPT showed significantly reduced cognitive engagement and less intellectual effort in transforming information into knowledge. Researchers at Psychology Today have drawn an even sharper distinction: adults who offload thinking to AI risk losing capacity they already built, while children who grow up relying on AI may never build that capacity at all. The implication is clear. If you use AI as a shortcut around the hard parts of learning, you end up with a fragile kind of competence that collapses the moment conditions change.

The productivity paradox

Here's a number that should make you pause: a 2025 field study found that developers using AI coding tools actually took 19% longer to complete tasks than those working without AI assistance. Both participants and experts had predicted significant speedups. Where did the time go? Screen recordings revealed that developers spent it prompting, interpreting, and correcting AI output. Without the knowledge to evaluate what the AI produced, the tool became a time sink rather than a time saver. This is the productivity paradox of AI. If you know your craft, AI removes friction and lets you move faster. If you don't, AI adds a new layer of complexity you're not equipped to manage.

The fundamentals are the moat

In a world where AI can generate first drafts of almost anything, the people who thrive will be the ones who can evaluate, refine, and extend what AI produces. That requires fundamentals. For a developer, that means understanding data structures, system design, and debugging. For a writer, that means knowing how to structure an argument and recognize when the logic doesn't hold. For a designer, that means understanding visual hierarchy and user behavior. For a strategist, that means knowing which questions to ask and which assumptions to challenge. These aren't skills you pick up by watching AI do the work. They come from doing the work yourself, struggling through the hard parts, and building the mental models that let you recognize patterns. As the World Economic Forum and others have argued, skills like creativity, judgment, adaptability, and critical thinking become true differentiators precisely because AI can handle the routine execution. The fundamentals aren't becoming less important. They're becoming the thing that determines whether AI makes you more capable or more dependent.

How to use AI well

None of this means you should avoid AI. It means you should approach it with the right mindset. Use AI to accelerate work you already understand. If you can evaluate the output, AI is a tremendous productivity boost. Use it to generate first drafts, explore alternatives, and handle repetitive tasks. But stay in the loop. Keep your critical eye engaged. Use AI to learn, not to skip learning. AI is an excellent tutor when you actively engage with its explanations. Ask it to break down concepts. Challenge its reasoning. Use it to fill gaps in your understanding, not to avoid building understanding altogether. Build the foundation first. If you're early in your career or learning a new domain, invest in the fundamentals before reaching for AI tools. The time you spend struggling with the basics is not wasted. It's building the very expertise that will make AI useful to you later. Know when to turn it off. Some work should be hard. The kind of work that rewires your brain and builds expertise requires genuine cognitive effort. If you outsource all of that effort to AI, you're trading long-term capability for short-term convenience.

The bottom line

AI is one of the most powerful tools we've ever had access to. But like any tool, its value depends entirely on the skill of the person wielding it. If you already know the fundamentals, AI will help you go faster than you ever thought possible. If you don't, it will give you the illusion of competence while quietly eroding the foundations you never built. The investment in fundamentals has never been more important. Not because AI is replacing skill, but because AI is making skill the multiplier that separates those who create real value from those who just generate output.

References