How to AI-proof your career?

How to AI-proof your career?

By getting good at skills where AI is naturally limited. I'll cover more in-depth about what that means below.

I agree that coding is not the only thing we do. Then what are other hard skills differentiator for a SWE?

Great question! Here's a couple great threads I highly recommend around that:

• "What are product skills and how to develop them in the age of ChatGpt and CopilotX?"
• "What are the Core skills a SDE should be consistently developing?"
• "What's the difference between a low performer software engineer and a high performer?"

Communication, System Design skills will get easier and easier to master given plethora of improved content on the internet.

While this is true, it's also true for literally any skill in existence as the amount of people teaching on the internet only grows and gets better over time. In the context of AI, which is what this post is about, I actually strongly disagree that AI will be a game-changer here, at least in our lifetimes.

The biggest buzz around tools like ChatGPT in the software world is that they can write code. However, the difference between code and something like communication is that the success criteria is far clearer and more binary for code (i.e. "Does it work or not?").

Of course, there are many ways to code a single task and code quality varies. But it's very easy to verify if code achieves the core objective: Does it build, deploy, and cover all defined use-cases? This is objectively provable.

However, with something like communication, it's way more fuzzy and this is where AI struggles. Let's take a very simple instance of communication: Asking a teammate a question. There is a huge range of outcomes:

1. Your question is confusing and terrible, so you don't get an answer. Your reputation tanks with your teammate and they don't answer any of your future questions.
2. Your question is okay, and you get an answer. However, it's not complete and your teammate isn't too excited to help you in the future.
3. Your question is solid, and you get a solid answer. Your reputation with your teammate remains neutral.
4. Your question is very good, being thorough and respectful. You get a great answer, and your teammate is more excited to answer your future questions. Your relationship with this person improves.
5. Your question and its presentation are so absolutely stellar that you not only get an incredible answer that teaches you a lot, your more senior teammate is so impressed that they decide to take you under their wing personally and mentor you 1 on 1.

I have seen all 5 of these outcomes happen across my career. The crazy thing is that the same input (i.e. your question content, body language, and display of gratitude) won't result in the same output all the time: Different people prefer different communication styles with different ideas resonating with them. What may be polite communication with someone might even be offensive to someone else! It is really hard for an AI to understand situations as nuanced and human as these.

These gray areas with effectively infinite skill ceilings are what make me love software so much and this is fundamentally what Taro is about: Deep-diving into this sea of complexity and nuance. I truly believe that AI won't be able to "solve" all this complexity in the next 50 years at least.

If you're interested in how to ask those "Level 5" questions, you should take my question asking course: [Course] Ask Great Questions That Get Great Answers Quickly

Should we start integrating AI in all our projects/workstreams in order to stay sharp?

Of course! AI takes the mundane software tasks like solving LeetCode problems and makes them 100% trivial. You should always be looking for opportunities to abstract away the more shallow work to focus on the deeper skills covered in the rest of this very response.

There's a viral tweet that lives rent-free in my head:

AI will not replace you. A person using AI will.

So the best thing to do is view AI as a tool, and figure out how to plug it into your workflows. I also think, as you imply, that anything related to human interaction (community groups, communication skills) will become more important.

The keyword I always come back to is leverage.

AI is taking over? Use that information to your advantage. Ask yourself these questions:

1. How can I leverage these new tools to create even more opportunities for myself?
2. What skills can I leverage to make myself more marketable or efficient at generating value for others?
3. What other non-AI-related tools can I leverage to force-multiply my output, whether that is with code or impact?
4. What are the areas in my life that I can improve that will raise my base competency, whereby these tools can then multiply (not all areas can be multiplied with software)?
   1. My health?
   2. My cognition through a better sleep schedule?
   3. My typing speed?
   4. My knowledge of IDE keyboard shortcuts?
   5. Design patterns?

Every little bit contributes to becoming AI-proof, but it is ultimately about standing out from the crowd and doing what the competition is too lazy or unaware to do.

Should we do start integrating AI in all our projects/workstreams in order to stay sharp?

Yes! Ever wonder how some people are so productive? Hint: it's mainly the tools they use and the systems and processes they've developed that work best for them.

Even better is to not just future proof yourself but others around you as well. I'm gonna help some other folks onboard with a modern devops solution they don't really know the tech behind and that's ok. We're just there to support each other and we move through it.

Some excellent answers here already.

I'll weigh in as "AI" practitioner.

AI is a marketing term that is not precise enough for what we need to answer your question, but I will roll with it so this doesn't turn into a 10 page essay.

Currently "AI" is mostly applied to the transformer-based models. These are completely devoid of true reasoning potential or ability to make complex connections. Those are major roles for engineers, so you are not competing with them and any company would be foolish to cut engineers because they have a more advanced auto-complete.

Adding reasoning to these models is something that we are actively working on as an industry, so some day it will be solved. Working "with AI" will be a very important skill.

To do that effectively we need two elements:

1. AI should be built with understanding of humans
2. You need to understand how AI works

Unless you are an ML researcher of engineer you really can't change much about 1, but 2 is within your control.

So to your question of should you be integrating AI into your workflows, the answer is absolutely yes.