NDNA 5 - From ‘It Works’ to ‘It’s Worth Using’: The AI Developer Divide

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:05

   Strawberry computer is the latest example of a movement that's rapidly gaining traction. Tools that promise to make building an app as simple as describing your idea. Powered by AI, Strawberry's vibe coding approach lets anyone generate working software in minutes. On the surface, it's an exciting evolution. Less friction, more access, faster results. But also forces us to confront a bigger question. What happens when generating code becomes

2. 0:28

   effortless? But building good products remains hard. This is NDNA number five. The truth is AI can write code, but it can't write judgment. It can't teach product taste, UX sensibility, or the deep contextual thinking that turns an app into something people trust and love. If you've seen a wave of AI generated projects lately, you might have noticed a certain sameness. Apps with identical layouts, familiar patterns, and a lack of distinct

3. 0:56

   personality. This isn't just an aesthetic issue. It signals a deeper risk. A homogenized software landscape where speed replaces craft. Now, if the lack of taste produces Dell products, the lack of security produces dangerous ones. The recent hack of the T app, which is a social platform where verified women could share experiences and warnings regarding men, shows what's actually at stake. While it's not

4. 1:22

   confirmed the app was AI built, it fits the pattern. low barrier to shipping but little evidence of strong security fundamentals. Security isn't just about code itself. It's about the architecture, deployment practices, data handling, and anticipating of bad actors. The faster and easier it becomes to launch an app, the more tempting it is to skip these steps. In a sense, we may be creating a new digital divide, not between those

5. 1:49

   who can code and those who can't, but those who can build securely and those who can't. Democratizing app creation will flood the world with more software than ever before. With outskilled refinement, most of it will be just insecure clones or novelty projects that fade away quickly. AI generated code is still dependent on the quality of the prompt and the judgment of the person using it. It can run, but runs isn't the same as safe, scalable, or even worth

6. 2:14

   using. The risk is not just low quality output and security. The risk is also stagnation in developer growth. If someone starts and stays at an average skill level, AI can help them ship more, but it doesn't necessarily make them better. Without learning the underlying craft, there's no clear path to progressing from average to extraordinary. The most successful products in this new environment won't

7. 2:37

   be just those that are a fastest to a prototype. They'll be the fastest of highquality prototypes, apps built by people who leverage AI to accelerate, but still bring strong design taste, deep domain knowledge, and rigorous testing. For a skilled developer, AI can be a multiplier. An above average developer can go from great and a great developer can become a 10x developer.

8. 3:03

   But below a certain certain threshold of skill, the multiplier effects flips and it actually results in more output, but no real improvement in quality or impact. If tools like Strawberry want to make a more sustainable, trustworthy products, they can't just stop at making deployment easy. They need to bake in security checks, enforce best practices, and guide users through safe, reliable deployment. Particularly in environments

9. 3:28

   like Neo's web 4 where code actually runs on chain. These guardrails shouldn't be about limiting creativity. It's more actually about how how it works and it works and it's safe. Without them, we'll keep seeing products that function in a demo but fail in the real world, including sometimes in ways that harm people using them. It's worth remembering that software built before AI wasn't automatically secure either.

10. 3:52

    But the scale changes the equation. When millions more people can generate and deploy code, and many of them lack the fundamentals, the probability of widespread vulnerabilities actually goes up. The challenge now isn't just teaching AI to write better code. It's ensuring the people using it learn how to build better products. The future will belong to those who can blend speed, security, and taste. If platforms like Strawberry can set that standard, AI built software can be both abundant

11. 4:16

    and excellent. If not, we maybe just be flooding the world with more of the same, only faster. This has been NDNA number five. If you enjoyed this episode, please like, subscribe, and share. Oh, heat, heat.