The Shifting Role of the Tech Stack

In 2026, the question of which tech stack to choose has changed fundamentally due to AI. While developers previously prioritized ecosystem size, syntax ergonomics, and the availability of tutorials, AI has flattened these advantages. AI models are increasingly capable of working with niche or brand-new libraries if provided with proper documentation and guidance. Consequently, the "best" stack is no longer dictated by market dominance but by the developer's ability to maintain and understand the resulting codebase.

The "Vibe Coding" Trap vs. Product Engineering

There is a distinction between "vibe coding"—generating throwaway software or internal tools where the stack is abstracted away—and building professional, monetized products. For the latter, the speaker argues that you cannot afford to be disconnected from your code. Blindly accepting AI-generated output leads to bloated bundles, unnecessary defensive programming, and potential security risks. The goal is to act as a "steerer" of AI, ensuring the code follows best practices and remains maintainable.

The Case for Simplicity and Reduced Dependencies

Supply chain security concerns have shifted the speaker's preference toward reducing external dependencies. Instead of relying on a library for every minor feature, the speaker now prefers:

  • Built-in Features: Utilizing native capabilities in runtimes like Bun or Node.js (e.g., built-in SQLite clients) to avoid unnecessary third-party packages.
  • Self-Building: Using AI to generate custom, lightweight implementations of features (like charting) that are tailored to the specific application rather than importing heavy, generic libraries.
  • Framework Consolidation: While meta-frameworks like Next.js are popular, the speaker emphasizes that full-stack frameworks (like AdonisJS or Laravel) offer a more integrated experience by including authentication, queues, and job scheduling out of the box.

Despite AI models having a strong bias toward React and Next.js, developers should not feel forced into these choices. The speaker notes that modern AI agents are capable of web search and can be pointed at documentation for any framework, including newer ones like TanStack Start. The decision should ultimately rest on what the developer is comfortable with and what allows them to maintain the highest quality of code.

Key Takeaways

  • Understand the Code: AI is a tool for speed, not a replacement for architectural knowledge. You must be able to read and debug what the AI produces.
  • Steer, Don't Just Prompt: Actively guide AI agents away from "defensive" coding patterns or unnecessary helper functions that bloat your application.
  • Prioritize Security: Reduce your dependency footprint by leveraging built-in features of modern runtimes like Bun or Node.js instead of importing external libraries.
  • Ignore the Hype: Don't feel pressured to use React or Next.js just because they are the default AI preferences; use frameworks that align with your project's needs and your personal comfort.
  • Focus on Maintenance: Choose tools that are well-maintained by trusted teams to ensure long-term security and performance.

Notable Quotes

  • "If you are building a product which you are monetizing, which you are selling, distributing, I believe it's important to understand the code and being able to understand the code."
  • "AI also... has a tendency to add lots of pointless helper functions into the code and be super defensive and try to have a fallback for a fallback for a fallback."
  • "I don't really want to use something which is not being updated at all anymore... I want it to be well maintained, of course, and that's like a bonus, but ideally by people I trust."
  • "Just because React is the favorite of AI models does not mean that you have to use it, I would say. I would use whatever you feel comfortable using."