The myths and what's actually happening

1. Four myths

If you read about AI and software engineering online, you'll see the same four claims repeated. Each one is partly right and badly stated. Stated cleanly, each one becomes useful.

Let's go through them.

2. Myth one — "AI will replace engineers"

What people mean: the model writes code, so the companies don't need people who write code, so engineers are obsolete.

What's happening: companies don't pay engineers to write code. They pay engineers to decide what to build, specify it clearly enough that something correct comes out, verify what was built, and own the result when it ships.

The model is excellent at writing code. It's mediocre at deciding what to build, weak at specifying without help, unable to verify against the real world, and incapable of owning anything. The parts of the job the model does well were never the whole job. They were the loud, visible part.

Two things follow. First, engineers are not being replaced. The work has shifted toward the parts the model can't do. Second, engineers who think their job was the loud, visible part are uncomfortable right now, and that's fair. The job they thought they had is smaller than the job they actually had, and the smaller part is the part that got automated.

3. Myth two — "Junior engineers are dead"

What people mean: companies aren't hiring juniors because the model can do junior work, so the entry-level door is closed.

What's happening: the work that used to be "junior work" got compressed. A 2015 junior spent their first six months learning to ship a small change end-to-end. A 2026 junior can ship a small change in their first week with the model's help. So the bar for what counts as junior moved up.

But the door isn't closed. It's shaped differently. Teams still need new engineers, because new engineers grow into senior engineers, and senior engineers leave. The hiring is just more selective on judgment from the start, and less selective on syntax. A 2026 junior is expected to specify a small task clearly, read the model's output, and catch obvious wrong things. That's a higher entry bar than it was, but it's not closed.

What's hardest is the very first job. Once you've shipped on any team, the second job is easier than it used to be. The first one is harder.

4. Myth three — "Now anyone can code"

What people mean: the model writes the code for you, so the skill of programming is open to everyone with no training.

What's partly right: yes, more people can produce working software now than ever before. People without engineering training can scaffold a working app, ship a small website, automate a personal workflow. That's real, and it's good.

What's not right: producing software and engineering software are different. Engineering is the part where the code keeps working after the team grows, after the requirements change, after the load goes up, after the original author leaves. Engineering is what makes the code maintainable, debuggable, and safe to change. Engineering is the discipline that pays off six months later, not the afternoon you ship.

The model lowered the bar for producing software. It didn't lower the bar for engineering it. If anything, the bar for engineering rose, because more software is being produced, and most of it is now maintained by someone who didn't write the first draft.

5. Myth four — "Senior engineers are obsolete"

What people mean: senior engineers got their seniority by writing a lot of code over many years. The model writes code now. So the long experience doesn't pay off anymore.

What's actually happening: the opposite. Senior engineers are more valuable than they were in 2015, and the gap between senior and junior is widening, not closing.

Why? Because the parts of the job the model can't do — judgment about what to build, taste about how to build it, the ability to spot the subtle wrong thing in a confident-looking draft, the experience to know which past mistakes apply here — those parts of the job are exactly what makes someone senior. The model amplifies whatever judgment is at the top of the loop. If the judgment is good, the model makes that engineer fast. If the judgment is bad, the model makes that engineer fast at being wrong.

A senior engineer is, more than ever, the floor of judgment on a team. The team can ship faster with the model. The team can also ship wrong things faster. The senior is what keeps the second number low.

6. The honest picture

Engineers aren't being replaced. The entry door is harder but open. More people can produce software, but engineering is harder, not easier. Seniors are more valuable, not less.

If you've heard the loud versions of these four claims and felt confused, you weren't wrong to be confused. The loud versions are mostly wrong. The boring, true version is the one above.

The next unit shows how teams are reshaping around this picture.