AI will not replace elite engineers

Summary

The discussion examines how AI affects software engineering, with a particular focus on database work, ORMs, performance, and the value of understanding systems beneath their abstractions.

Engineering and critical thinking

Hussein Nasser describes programming and troubleshooting as rewarding partly because engineers work through problems themselves. He worries that relying too heavily on AI can reduce that learning and weaken critical thinking. AI may speed up certain tasks, but skilled engineers still need to understand how their systems behave.

Database design and tuning

Database work involves much more than designing normalized tables. Engineers must choose suitable data types and indexes, and tune queries for the particular database platform. Nasser contrasts composite indexes with indexes that include extra columns, noting that database systems implement these features differently.

PostgreSQL performance

Nasser’s assessment is that PostgreSQL offers less control over query plans than platforms such as Oracle and SQL Server, though he also notes that it continues to improve. He highlights PostgreSQL 18’s asynchronous I/O support, including io_uring, as a useful newer feature.

He argues that application developers sometimes have context the database lacks—for example, knowing a table will become large—and would like more ability to guide the planner with hints.

Migration and troubleshooting

Moving data between databases requires attention to type differences and rebuilding indexes. Nasser recounts a performance problem on Amazon Aurora PostgreSQL: a workflow that ran quickly locally saturated CPU in Aurora. He traced the issue to stale statistics, which led PostgreSQL to choose full-table scans rather than index-based plans. Running VACUUM ANALYZE refreshed the statistics and corrected the query plans.

The example illustrates why engineers should inspect execution plans and understand database maintenance.

ORM trade-offs

Nasser views ORMs as useful for making code more approachable, especially in team settings, but warns that they can conceal expensive database behavior. For example, loading objects one at a time can trigger many queries. Retrieving large result sets and sorting them in application code can also waste database CPU, backend memory, and network bandwidth.

He recommends reviewing what an ORM is actually doing and pushing suitable filtering and sorting into the database. Raw SQL can hide problems too—for example, through SELECT *—so the broader issue is hidden complexity, not only ORMs.

Scalability and security

The speakers discuss how inefficient ORM usage can burden application servers and databases, particularly as concurrent usage grows.

On security, Nasser emphasizes limiting access to fields a system does not need and enabling TLS for database connections. These concerns apply to both ORM-based and handwritten SQL.

Practical use of AI

Nasser uses AI as a faster reference tool—for example, to find a query for inspecting locks in an unfamiliar database or to recall JavaScript reduce and map syntax. He is reluctant to have AI generate tests because he wants to understand and be able to debug them himself.

His approach is selective: use AI for bounded information needs, but be cautious when the output is central to correctness or depends on project-specific context.

Planned learning material

Nasser says he is developing a learning platform with intermediate-to-advanced courses and technical deep dives into systems such as Nginx, HAProxy, and databases. He also describes a planned site for analyzing important bugs and features, and a book about backend communication patterns—from incoming connections and request handling to worker pools and database connection pools.

Reviews, guides, or tutorials

No product review or step-by-step tutorial is presented. The discussion includes a practical PostgreSQL troubleshooting example and outlines planned advanced courses and technical explainers.

Main speakers

• Hussein Nasser — Backend engineer, author, and host; discusses databases, backend performance, ORMs, and AI-assisted development.
• Rashed Hasan (also rendered as Rashid Al-Hasan in the subtitles) — University of Nebraska graduate student and interviewer; asks about database design, migrations, ORMs, scalability, security, and AI.