About this video

• Video Title: What I had to unlearn to grow in tech (in the AI era)
• Channel: My 50 cents
• Speakers: Nimisha Aigiri, Ricardo Ferrer, Vive Ka
• Duration: 13:12

Overview

This video discusses the importance of unlearning old habits and mindsets to adapt and grow in the current AI era of technology. It highlights how AI is changing the landscape of software development, emphasizing the need for new approaches, skills, and a shift in perspective, particularly regarding collaboration with AI tools rather than seeing them as mere replacements. The speaker shares personal experiences and insights gained from attending the InfoQ Dev Summit in Boston.

Key takeaways

• The Need for Unlearning: The core message is that growth in the tech industry, especially in the AI era, requires unlearning outdated knowledge, habits, and mindsets rather than just accumulating new information.
• AI as a Collaborator: AI is presented not just as a tool but as a collaborator, changing how we learn, code, and develop software. Examples like Perplexity and code assistants illustrate this shift.
• Human Judgment Remains Crucial: Despite AI's capabilities, human judgment, experience, and craftsmanship are still essential for creating quality, lasting software. AI struggles with understanding context, intention, and socio-technical aspects.
• Evolving Developer Roles: The roles of developers are evolving. Prompting and reviewing AI-generated code are becoming key skills. Senior engineers may focus more on coaching and mentoring.
• Importance of Foundational Skills and Practices: Core principles of software architecture and engineering remain valid. Strong engineering practices like code reviews and building fundamental knowledge before relying on AI assistants are more important than ever.
• Value of Community and Conferences: In the fast-paced AI world, community knowledge and conferences serve as valuable platforms for accelerating learning, sharing experiences, and discovering practical patterns, thus compressing learning time.

In the video, unlearning examples are discussed in the context of adapting to the AI era in technology. Here are the key examples presented:

• Traditional "Know Everything" Mindset vs. AI Collaboration:

  • What to unlearn: The old way of growing in tech involved knowing as much as possible, deeply and broadly, by staying late, reading documentation front to back, and debugging for hours without immediate answers. Communication was primarily asynchronous and slow.
  • What to learn/adopt: Now, the focus is shifting. Instead of trying to know everything, developers need to learn how to effectively collaborate with AI. This means understanding how to prompt AI, guide it, and critically review its output, rather than solely relying on individual deep knowledge.

• Relying Solely on Manual Coding vs. Using AI Assistants:

  • What to unlearn: The habit of writing every line of code manually, believing that deep individual coding is the primary path to success.
  • What to learn/adopt: Embracing AI code assistants as collaborators. This involves learning to instruct them, understand their limitations (like not reading minds or losing context), and integrating their output thoughtfully into the development process.

• Assuming AI Understands Context and Intent:

  • What to unlearn: The belief that AI, especially large language models (LLMs), inherently understands the nuances of context, intention, or complex decision-making in the way a human does.
  • What to learn/adopt: Recognizing that developers must explicitly provide context and guide the AI. This led to the practice of using markdown files to document design decisions, goals, and constraints for the AI, helping it maintain context over longer sessions.

• Lack of Explicit Design Documentation for AI:

  • What to unlearn: The past practice of keeping design decisions implicitly in one's head or in scattered notes, not necessarily formalizing them for easy recall or sharing.
  • What to learn/adopt: Actively documenting design decisions, goals, and constraints in a structured way (like a markdown file) to provide clear context for AI collaborators, especially when they lose track.

• Refactoring as a Standard Practice:

  • What to unlearn: The expectation that AI-generated code will always be perfect and immediately usable without significant human intervention or refinement. Studies show a significant amount of AI-generated code gets deleted.
  • What to learn/adopt: Understanding that while AI accelerates coding, human judgment is still needed to ensure the code is "worth keeping." This means continued emphasis on code quality, design, and potentially still needing to refactor or carefully integrate AI suggestions.

• Viewing AI as a Replacement for Fundamental Skills:

  • What to unlearn: The idea that AI will make fundamental programming languages, design patterns, and software architecture knowledge obsolete.
  • What to learn/adopt: Recognizing that these foundational skills are still critical. AI can assist, but understanding the underlying principles is necessary to effectively guide AI, design reliable and secure systems, and make sound architectural decisions.

• Over-reliance on AI for Complex Socio-Technical Decisions:

  • What to unlearn: Trusting AI to make intricate decisions about system design, service granularity, team dynamics, or aligning technology with human factors.
  • What to learn/adopt: Understanding that these complex, often "lived" and undocumented, socio-technical aspects require human experience, judgment, and intuition that AI currently cannot replicate.

In essence, the video advocates for a proactive and critical approach to integrating AI, which involves questioning existing assumptions and adapting one's own practices and understanding.