AINDA VALE A PENA SER PROGRAMADOR NA ERA DA IA?

Summary: AI and the Future of Software Engineering

• Core question: Whether software engineering still makes sense as a career in the age of AI—given concerns about automation “writing code, tests, documentation, and doing code reviews.”

Market vs. Profession

• The speaker argues that a slower, more selective job market (layoffs, fewer hires, more competition) is often mistaken for the end of the profession.
• The video claims the hiring boom during 2019–2022 was abnormal—tied to pandemic-era capital and hiring surges—and shouldn’t be treated as the baseline.

What AI Changes in Developer Work

AI is expected to increasingly handle implementation-level / operational tasks, such as:

• CRUD endpoints
• boilerplate code
• repetitive code generation

As a result, the role shifts toward higher-level responsibilities, including:

• problem understanding
• solution design
• risk analysis, trade-offs, and architecture choices
• handling scale and large applications
• working with large data volumes
• validating that the solution delivers the intended business outcomes

“Know the When and Why,” Not Just the What

• The speaker emphasizes that even if AI can recognize patterns and techniques, developers must know:
  • when to apply them
  • which edge cases might be missing from prompts
• A key warning: if a developer doesn’t even have the relevant problem on their mental map, AI won’t include it—leading to production failures.

Distributed Systems Example: Dual Write Risk

A concrete scenario illustrates why AI won’t “infer” everything correctly without developer knowledge:

Example flow

1. Create an endpoint that writes a user to Postgres
2. Publish an event to a queue such as Kafka or RabbitMQ

Failure case

• If message publishing fails after the DB insert:
  • data is written to the DB, but the message is never delivered
  • other services remain unaware
  • the system becomes inconsistent
• This is described as the Dual Write problem.

Mitigations (requires real engineering knowledge)

• Transaction Outbox Pattern
• retries / retry techniques
• deduplication
• idempotence

Why “You Don’t Study Because AI Knows” Is Dangerous

The speaker distinguishes between:

• not knowing a solution, versus
• not knowing that a problem exists at all

If something isn’t on your mental map, it won’t appear in prompts—and AI won’t generate the missing parts of a robust solution.

Trade-offs Can’t Be Solved by Prompts

The video argues you can’t simply request “all best practices” or “all design patterns” and expect correct correctness/performance outcomes.

Architecture decisions involve trade-offs, for example:

• Strong consistency / ACID
  • more correctness, but higher latency and unavailability risk
• Eventual consistency
  • better latency/availability, but requires careful handling of inconsistent data
• Optimistic vs. pessimistic locking
  • prevents race conditions, but can impact performance
• Monolith vs microservices
  • microservices can increase independence, but introduce network partitions, latency, observability needs, partial failures, and CAP-theorem complexity

Professional Differentiation in an AI World

• Two developers using the same AI can produce different results.
• The key differentiator is the ability to evaluate requirements and constraints, such as:
  • atomicity needs
  • whether ACID is required
  • race conditions to consider
  • the appropriate consistency model
• In this view, the “value” shifts toward directing AI execution so the system behaves correctly in the real world.

Learning Emphasis and Referenced Guides

• The speaker repeatedly emphasizes learning fundamentals rather than chasing hype/tools.
• Mentions an “Architect’s Roadmap” career guide (linked in the description) that outlines a pathway:
  • zero → junior → full → senior → software/solutions architect
  • including guidance on studying fundamentals in the right order
• Additional referenced topics/videos include:
  • Dual Write
  • idempotence
  • advanced data-query approaches (e.g., using Full Text Search instead of a naive LIKE)

Main Speaker / Source

• Renato (referred to as “Renato” throughout; primary presenter on “this channel”).