I joined my first company as a trainee integration engineer. I was employee number 30. I stayed 14 years, got promoted seven times, became an application architect, and hired more than 80 engineers along the way.

I learned by being stuck

Nobody handed me a curriculum. I taught myself .NET, then SQL, then Oracle, mostly because the work in front of me needed it and nobody was free to do it for me.

Somewhere in those years, coleagues started calling me tech-agnostic. I took it as a compliment. I also knew what it described: years of being the person who didn't know something yet and had to figure it out before the deadline.

My first real application came when MVC was just catching on in ASP.NET. I built an internal tool with it, and almost everything was new to me: the pattern, the backend, the frontend, the database. I learned it by sitting with seniors, living on Stack Overflow, and putting in hours after work because I was hungry to learn. When ASP.NET Core 1.0 arrived, I did it all again. Then Angular. Then Node. Each time, I learned the most in the hours I spent stuck.

Most of my judgment came from that slow, uncomfortable stretch. When I review a design today, I'm drawing on hundreds of small failures I worked through myself. Queries that crawled. Integrations that broke in ways the documentation never mentioned. I can't point to the day I became senior. I can only point to the pile of things I had to untangle.

I can't point to the day I became senior. I can only point to the pile of things I had to untangle.

The training ground is shrinking

Today I'm a VP of Engineering, and I still write code. Plenty of what took me an afternoon as a trainee, an AI tool now drafts in minutes. That speed is what worries me.

The work AI does best is the work juniors used to learn on: boilerplate, first-pass queries, glue code between two systems, the bug that's obvious once you've seen it three times. None of it was glamorous. It was where you wrote the clumsy version, a senior showed you why it would break, and next time you saw it coming.

A junior engineer in 2026 can skip most of that. They prompt, get working code, and ship. The code often looks fine and isn't. The Stack Overflow 2026 Developer Survey found 46% of daily AI users had at least one production incident in six months traced to code they hadn't closely read.

The code problem has a fix. Better review, better gates. The harder question for me is about the person. If you never write the clumsy version, do you ever learn to spot it?

If you never write the clumsy version, do you ever learn to spot it?

So I'm asking

I don't have a clean answer. When I think about hiring now, I care less about how fast someone produces code. I care more about whether they can explain why it works, and what they do when the AI is confidently wrong. I built those habits by being stuck. I'm not sure how a new engineer builds them when the tool rarely leaves them stuck.

Maybe the path moves. Juniors might learn by reviewing AI output, or by debugging what the agent got wrong. The senior engineers of 2035 might have a kind of judgment I can't picture yet. I'd like to believe that. I haven't seen anyone design that training path on purpose.

So I'll ask the people living it.

If you're early in your career: where are you building judgment that AI can't hand you?

If you lead engineers: what are you doing on purpose to grow seniors, now that the old training ground is shrinking?

Key takeaways

  • Senior judgment comes from problems you worked through yourself, not from code you shipped.
  • Every new stack (ASP.NET MVC, ASP.NET Core, Angular, Node) was learned the same way: building, getting stuck, and working through it.
  • AI tools now do the routine work juniors used to learn on.
  • AI-written code can look correct and still fail in production when nobody reads it closely.
  • In hiring, explaining why code works matters more than producing it fast.
  • Few companies have deliberately designed how juniors build judgment in an AI-first workflow.

Common questions

How is AI changing the path for junior engineers?

AI coding tools now handle boilerplate, first-pass queries and glue code. That was the work juniors learned on. They can now ship working code without writing the clumsy first version that used to teach them what breaks.

How do software engineers become senior?

In Ravi Bhushan Ojha's experience, through problems worked through firsthand. He started as a trainee integration engineer and learned ASP.NET MVC, ASP.NET Core, Angular and Node by building with each one, sitting with seniors, and working after hours until things made sense.

What should engineering leaders look for when hiring juniors in 2026?

Ravi looks less at how fast someone produces code and more at whether they can explain why it works and what they do when an AI tool is confidently wrong.

Can junior engineers still build judgment using AI tools?

Possibly, by reviewing AI output and debugging what agents get wrong. Ravi's concern is that few companies have designed that training path on purpose.

Who is Ravi Bhushan Ojha?

Ravi Bhushan Ojha is VP of Product Engineering at Zithara Technology, an AI-first CRM for retail brands, with 15+ years in product engineering. He spent 14 years in life sciences software, growing from individual contributor to engineering leadership.