Korva is a scroll of the internet’s best ideas.

If agents are trusted to know every language and write all the code, nobody can tell whether a product is good or bad, or whether it is working as expected. Netflix takes a lot of risk, fails fast, and recovers fast, and that depends on understanding how the systems work. An agent can help detect and triage an anomaly faster, but a person still needs fluency in what is being built and how it works, to know if it is good and how to fix it.

Code that models and agents write today can be very hard to follow. Performance may improve with no explanation of why, and if the thing breaks, nobody knows how to fix it. That looks unfamiliar and unsettling at first glance, and the learning curve is steep right now, since the tests and reasoning needed to get comfortable with it are still being worked out. Engineering will likely evolve to be at ease with that and to guide agents so the output can be trusted.

Engineering has shifted dramatically before. Over the last 10 or 20 years it changed just as much, much as filmmaking has changed over the last 100 years with new tools. The cycle seems to be speeding up.

Did compilers once raise the same worry?

John Backus led an IBM team that delivered the first FORTRAN compiler in 1957. Many programmers at the time distrusted automatic translation and believed hand-written machine code would always run faster. Backus's team spent years making the compiler's output efficient enough to answer that objection. Within a few years, high-level languages were standard for scientific computing.

Why does hidden complexity cause trouble?

Software developer Joel Spolsky coined the Law of Leaky Abstractions in a 2002 essay. It states that all non-trivial abstractions leak to some degree, which means the details underneath eventually show through. When something breaks, the person fixing it needs to understand the layer that was supposed to be hidden. Spolsky used examples from networking, databases, and programming languages.

How has filmmaking technology changed over a century?

The Jazz Singer, released in 1927, was the first feature film with synchronized dialogue sequences, and it ended the dominance of silent pictures within a few years. Color became standard, and Jurassic Park in 1993 showed audiences photoreal computer-generated animals. Toy Story, released in 1995, was the first feature film made entirely with computer animation. Digital cameras and projection then replaced most film stock in theaters.

More from Lenny's Podcast: Product | Growth | Career

Work & LeadershipAI fluency at every levelNetflix expects AI fluency from everyone, senior leaders included, and lets candidates use AI tools in coding interviews.
Enjoyed it?