Engineering managers who build are pulling ahead

Estimated reading time: 5 minutes

Key takeaways:

• Engineering managers now need hands-on building experience to lead effectively.
• Taste (and the ability to spot AI slop) has become the critical review skill.
• The old speed-vs-quality tradeoff is disappearing, so quality is no longer the cost of moving fast.

When the AI revolution is discussed, the headlines are usually focused on productivity gains for writing code, changes in hiring engineers, and the challenges faced by non-technical team members. The impact on engineering managers is rarely discussed widely.

The skills required to be an effective manager are now completely different from what they were even a year ago. Here I will share my perspective on what has helped me be effective in this day and age. The answer is pretty straightforward: building things.

This is arguably the best time to build software in human history, as going from an idea to a working prototype is now a matter of hours, not weeks. If you’re the kind of manager who thinks they’re too rusty to write code, I’ve got news for you: you won’t be writing a single line yourself anyway.

You might ask, why is this important, and why should you care? Well, the paradigm has completely shifted. Things like context windows, harnesses, embeddings, vector databases, MCPs, (and so much more) are concepts that only came into existence in the recent past. Without knowing each of these deeply, you won’t be able to lead engineers who are supposed to build with them.

Below are the lessons I’ve learned exclusively by building side projects (both at work, and outside of work) that have marginally improved mine, and more importantly, my team’s output.

Taste has never been more important

What makes us human? What makes AI not? While these may be deeply philosophical questions, the answer I’ve found is, taste. Primarily, the ability to recognize AI slop. This intuition is only developed through getting your hands dirty, deep into three hours of vibe coding at 1am, wrestling with AI tools after realizing they can’t one-shot something that you will like.

Whether you use generative AI for building applications, or writing documents, the first version is always going to look very AI-generated (not a compliment). I’m not even talking about the em-dashes, the “its not X, its Y” sentences, websites with the same hue of purple, or design docs littered with emojis.

It runs deeper than that. You end up with something that looks perfect, but is far from it. It’s never entirely accurate, and more importantly, it’s not you. The primary reason for this is, no matter how detailed your prompting game is, is that you will never be able to type the many many layers of your brain into a text box.

AI is designed to find the median of human taste, to condense everything to a common plane of creativity, which is precisely what makes its output so easy to spot.

Reviewing code doesn’t have to be a drag

I lead a team that builds financial systems at YouTube, ensuring music royalties and media streaming revenue gets paid out on time and accurately, globally, each month. Billions of dollars move through the code we write and maintain. This means that any bug in the system has the potential of becoming a financial event that can impact creators directly. In this world, there is very little room for error. 

Reviewing new code that is increasingly generated using AI is a new skill in itself. As output is scaling, reviewing is ever important, without it necessarily becoming a bottleneck.

For reviewing complicated changes, I discuss alternatives and tradeoffs with AI, poke holes in the implementation, until I’m convinced I understand 100% of what is being pushed. As a manager, I’m naturally one degree away from the weeds of things. This helps me be in tune with each change that goes through, expanding my own understanding of the codebase. It actually makes reviewing code a fun exercise, and the team’s output speaks for itself.

You don’t need to take shortcuts anymore

Every well managed project consists of a planned timeline, implementation milestones, and a well defined scope. In the past, timelines always had a healthy tension with the scope. You often had to choose: cut features to meet a deadline, or sacrifice code quality to squeeze in functionality.

Today, that friction is effectively gone because the time and effort for creation has reduced significantly and we no longer have to settle for just ‘good enough.’ A plan that relies on future technical debt as a strategy is not a good plan. We can now build things the right way from day one, tackling the hard, foundational problems first rather than leaving them for a mythical future refactoring sprint.

This shifts the mandate from shipping as fast as possible to shipping as thoughtfully as possible. The key unlock is the ability to fail fast with prototypes, validating real world feedback sooner, and pivoting if needed. We can maximize creative output, explore the ambitious ideas we’ve always shelved, and build products with architectural integrity that used to take months of planning. When the cost of doing it right is just as low as the cost of cutting corners, the only remaining constraint is your own ambition.

For my team, this has meant automating workflows that handle hundreds of millions of dollars in annual royalty payouts, reducing manual overhead by thousands of hours and achieving granular accuracy that was once considered impossible.

New York • September 15 & 16, 2026

The pace of change keeps accelerating. See what other leaders are doing about it, at LDX3 New York.

Explore

The future of AI-native leadership

We are in a new era where the distinction between a manager and a builder is already rapidly dissolving, and we should embrace it.

For me, the manager-builder model has become necessary for maintaining the right level of technical intimacy that allows leading with confidence, spotting differences between AI slop and true innovation, and ultimately advocating for the hard, foundational work that drives long-term value.

So if you are a manager reading this, understand the machinery and embrace the fact that you no longer have to settle. The cost of doing it the “right way” has never been lower, and along with it, the potential for impact has never been higher.