As the software industry continues to evolve, new tools, frameworks, and specs keep emerging. At first glance, this indicates constant change and diversity. Historically, software has evolved rapidly, and AI is accelerating that process. But generative AI can be a double-edged sword, both raising the waterline and diminishing ingenuity.

By automating repetitive tasks, developers can focus on outcomes, explore new ideas, and enhance accessibility, performance, and richer user experiences. However, in practice, automation often amplifies existing patterns. The same tools that enable efficiency also reinforce defaults, entrenching today’s best practices deeper into workflows, validation, and design decisions.

Why “AI’s Favorite Color” is Purple

I asked ChatGPT, and this is what it said: The color purple appears so often in generated code when a model aims to be distinct yet safe. It is less aggressive than red, less corporate than blue, and not as loud as neon. When an AI is tasked with designing something appealing, purple tends to fall near the middle of the probability curve. The models were trained on, including Tailwind and other frameworks, which also influence the current output. It is not fair to say AI only provides the next statistically relevant token, but it does come with its own bias.

Bootstrap Déjà Vu

UI Frameworks, particularly Bootstrap, didn’t just provide components; they fundamentally shaped the web’s appearance for a decade. Shaping user interfaces by reducing friction, incorporating best practices, and prioritizing speed over careful consideration. Bootstrap is optimized for defaults, and defaults tend to scale more quickly than personal preferences.

Generative AI is similar to Bootstrap’s default styles. Just as Bootstrap offers standard design language, AI provides statistical averages based on millions of data points. While Bootstrap didn’t eliminate designers, it lead to a homogenization of design choices. Many sites began to look the same; you could pick a Bootstrap site out of a lineup.

Coding standards will increasingly solidify not solely because of collective respect for established practices, but also because current models are trained on them and continually reinforce them. At scale, this creates an echo chamber where familiar approaches are repeatedly produced and validated. Meaningful deviations become rarer, not because alternative ideas lack merit, but because they fall outside the model’s learned center of gravity, which is used by the masses.

Agency and Intent Matter More

There will always be those who challenge entrenched norms. These individuals and teams often act as catalysts, introducing new patterns and reframing problem-solving approaches. Their work is significant and shapes the field’s direction. With AI hijacking the developer life cycle, innovation could be hindered by entrenched norms, especially when competing with systems optimized for consistency, predictability, and repetition.

To elevate AI-generated code beyond the baseline level, we must provide meaningful direction and insights. While generating code with AI is straightforward, validating it and providing well-structured, detailed instructions are crucial to adding the human element. Which I argue is passion. Although it’s impractical to read every piece of generated code, it’s essential to review the reports and analyses of the code being produced. Utilize unit testing effectively to validate that the generated code works well for both humans and AI. Take the time to read the planning prompts and review the output so you know what the AI is doing and why. It’s tempting to take a back seat when developing with AI, but doing so would mean using it as a crutch rather than a tool. Your awareness and involvement are essential. Don’t be afraid to get your hands dirty in the process.