What makes a senior in the AI-driven era

For the past year, I have been exposed to a lot of news and content related to AI (who doesn't?).

Did I get anxious about it? Yes, to a certain extent.

You can't really hide yourself from all the AI stuff. It's everywhere, and not trying to get to know it would probably just be increasing your anxiety. In tech, agentic coding is especially popular, and it is not hard to see why. I started giving it a try for the past month, and it got me thinking: if AI can do so much, what does it even mean to be a senior anymore?

Let me start by saying that my tinkering with agentic coding is not even close to some of the crazy stuff I have been seeing people do online. I did not set up a personal assistant to which I give access to my calendar, email, or anything. I did not try running a swarm of agents to do work in parallel. My tinkering has been much simpler than that.

1. I brainstorm on claude.ai to come up with a plan
2. I use Claude Code to implement the tasks

I started with personal projects, where it's less risky, and there's more room for experimentation.

If you have a well-defined problem and a clear solution in mind, AI can be a powerful tool to help you execute it quickly and efficiently. It can scaffold, implement, refactor, and explain code at a pace that can significantly speed up the development process. This is especially true for tasks that are repetitive or have well-established patterns.

Skills that are typically valued in a senior engineer, such as establishing conventions, drafting up plans, and breaking down problems, are still very valuable. Things that you usually do for your less-senior teammates, you still need to do for the AI.

To be honest, a lot of the work you have been doing as a software engineer is probably a solved problem. Either you already have a good idea on how to solve it, or you can do some research and find out how to solve it. Even during this phase, AI could help a lot by helping you research faster. Think about it as discussing with a colleague. Sure, AI can hallucinate, but humans aren't always right either. At the end of the day, you will be the decision maker and as such you should sense check the plan anyway; AI is just a tool to help you get it done.

Sounds engagement-baiting, but it's true.

Most parts in our software aren't unique. That's why we use reusable libraries, copy-paste from StackOverflow, and that's why there are a lot of xxx-starter repos on GitHub.

This is why AI can be so good at "one-shotting" apps. These are solved problems. Though the larger the scope, the less chance of it at one-shotting it. Context window is still a bottleneck, as it is the case with humans' cognitive load capacity.

Most engineers, probably at some point in their career, have thought to themselves: "It would be so much easier to work on a greenfield project, starting from scratch without the legacy stuff".

It's kind of the same thing with AI. If you allow them to start from scratch, there are fewer opportunities to break something. You can also scaffold a project yourself, but it probably won't be as fast.

With new projects, you can establish your own conventions, document everything, and make sure everything is neatly organized. Long-running projects suffer from technical debt, outdated documentation, and maybe even some obscure knowledge that only a few people in the team have.

Reflecting on my personal career, seniority seems to have been more about the decision-making and judgment around the work, rather than the raw execution of it. We probably know someone at work that we know we can always count on when we have a question that we aren't sure about (or maybe you're that person!). What makes them senior is not that they can write code fast, but bouncing ideas off them or chatting with them always leads to better clarity.

Seniors aren't necessarily the ones who always have the answer, but collaborating with them almost always leads to productive results. Some of the traits I find valuable in seniors are:

• They ask good questions
• They provide meaningful feedbacks
• They understand tradeoffs and can articulate them well
• They understand the team and product situations
• They are trustworthy

Not all of these traits are replicable by AI (at least not yet!). Things that make you valuable as a senior still do. Traits of a good tech lead are, in general, still very valuable.

AI does help you execute faster. Faster iterations lead to faster learning, which is a good thing! The bottleneck is not writing code; it's having the necessary context about the whole picture to decide what to write and why.

I have been thinking about empathy; in a world where things are becoming more automated, a human touch stands out even more. The engineers that I look up to are usually the ones who are not only technically competent, but also empathetic and good at communicating with others. They understand the human side of software development, and they know how to work well with their teammates.

Good judgment and taste become even more distinct. Sure, AI can explore design directions as well, but at the end of the day, you are still the one making the calls. You are the one who decides whether or not something is good or not. Understand the tradeoffs, make the call, and own the decision.

Learning was and still is a valuable skill to have. AI helps you learn faster. Instead of hours of googling, you can get a more targeted answer quicker. "But AI can hallucinate and be over-confident", you say. Yes, but searching on your own could also lead you to inaccurate answers or outdated information. At the end of the day, you should verify and understand the solution you get anyway.

Proper documentation, establishing good conventions, and automated tests and CIs are more important than ever. They help both humans and AIs. As the speed of outputting code increases, things that prevent the wrong code from getting into production matter more.

The floor has been raised for everyone, including juniors. In 2017, simply setting up a full-stack web-app could get you a lot of entry-level job offers. Now, it can be done in a few minutes with AI, and thus it's no longer a good signal for hiring.

I believe that the things worth investing in as a senior are also worth investing in as a junior. Unfortunately, you can't improve in some aspects that are only accessible if you have a real-world project.

In the past, a lot of people couldn't start learning because they:

• didn't have anyone to ask questions to
• didn't know where to start

Some more proactive people would try cold-reaching more senior folks in the industry to ask for advice, though it doesn't always work out. Nowadays, AI can be your 24/7 learning partner. The goalpost for juniors is farther away, but the tools to reach it are available. I remember the days when I felt hopelessly stuck on trivial problems. StackOverflow answers didn't seem to make sense. Asking people on the internet seldom got me a response. These are less of a problem today.

Like most things in life, things change. Your game plan needs to adapt. If you are a self-starter, proactive individual, you will go far.

I honestly don't know what to expect from the future, considering how rapidly things are evolving. But one of my biggest hopes is that we find more ways to connect with our human side, and that a world with more AI somehow leaves more room for that, not less.