How to Survive the AI Era as a Software Developer

If you’re a software developer, you’re probably worried about everything happening with AI and LLMs. In this video, I’ll cover 7 things you can do as a software engineer to help you not only survive this avalanche of AI tools but maybe even thrive in it.

🔖 Chapters:
0:00 Intro
1:14 Lean into AI
2:50 Use AI to learn more quickly
3:51 Specialize
4:40 Become more interdisciplinary
5:41 Be innovative
6:30 Better understand security and privacy
7:14 Develop Soft Skills
7:58 Final thoughts

#arjancodes #softwaredesign #python

[snarkyboojum]

Quick summary using AI -

1. Embracing and integrating AI tools into the workflow: The speaker emphasizes the importance of software engineers embracing AI tools and integrating them into their workflow to not only survive but thrive in the changing landscape. This includes using AI to learn more quickly and stay updated on industry changes.

2. Specializing in areas where AI cannot replace human expertise: The speaker suggests that software engineers should specialize in a particular area of software engineering where AI cannot yet replace human expertise. This specialization can help them maintain their value and relevance in the face of AI advancements.

3. Gaining knowledge in other disciplines: The speaker encourages software engineers to gain knowledge in other disciplines such as business or psychology. AI is not yet proficient in understanding certain aspects like user experience, so having expertise in these areas can provide a competitive advantage.

Overall, the video emphasizes the need for software engineers to adapt to the presence of AI and large language models by embracing AI tools, specializing in areas where human expertise is still crucial, and gaining knowledge in other disciplines to enhance their value and capabilities.

[elau]

AI/ChatGPT is a tool. I see it as a better Google search engine for programmers to look stuff up with context.
Tech is a very fast moving industry where things get obsolete very quickly. It is not tax preparation where you skills have a longer shelf life. One need to keep up with the changes for your next gig is asking for new tech stack and skills.
It is not easier when there are other things in life that gets in the way but the one constant skill is the skill to continue to learn.

[digiryde]

@ArjanCodes - This is an excellent overview of what can be done!

In my opinion, the most important skill with LLMs is designing work flow. Looking at all of the successful prompts, they are all about taking a process you outline to them and returning the skeleton code for you to flesh out. This is really no different than writing requirements for a project to hand off the development team. I find that this is actually a hard skill for most people to learn.

Basically, at its most basic, this is conceptual translation in much the same way that people translate one language to another, but can be much more difficult due to the nuances that have to be captured and then expressed in a way that the LLM can work with.

Also, you need to know how to code in the first place so you can catch the issues that come up in the returned code. Some of those may be easy to spot, such as the code will not run, while others may be very difficult to spot as they lead you down dark alleys of code that suddenly dead end.

Which leads to one needs to exercise being able to model data and systems in their head. If you can not see the model, how will you know the model works? Like planning, architecting, and then constructing buildings, the bigger they get, the easier it is to screw the whole thing up in some part you did not quite "understand" or perceive as it would be.

[CodeBadger-xhxg]

Good advice! We should focus on what we can control...in this case, our ability to adapt.

[Zale]

00:00 Arjan addresses concerns about AI and large language models replacing software engineers and suggests ways to survive and thrive in the AI era.
00:55 Instead of fighting AI, software engineers should embrace and learn about AI tools to improve their workflow and create higher quality software.
02:19 AI tools are helpful in quickly generating boilerplate code, but it is important to think through the overall design first.
03:13 Using AI can help software engineers learn more quickly and stay updated on industry changes, increasing their chances of success in the job market.
04:12 Specializing in a particular area of software engineering or gaining knowledge in other disciplines can provide value that AI cannot replicate, such as cybersecurity or UX design.
05:38 Software engineers can choose to specialize in a specific subject or become interdisciplinary to gain a broader perspective.
06:05 Focusing on innovation and creative thinking can set software engineers apart from AI, as AI struggles with these areas.
06:34 Understanding security and privacy concerns, especially on the business side, can be valuable as businesses are cautious about data sent to proprietary AI models.
07:30 Developing soft skills like communication, leadership, empathy, and teamwork is important as these skills are difficult for AI to replicate and are valuable in interacting with customers and stakeholders.
08:28 Arjan encourages software engineers to work alongside AI technologies and use them to enhance their own capabilities and productivity.

[lien-chinwei]

I agree with that AI is very useful summarizing and synthesizing information. Everyday, I try to use AI to help me learn new knowledge!

[Personal-M.I.S.]

I'm going to disagree with jokwon that "coding is 10-20% of an engineer's task". I run a small dev company of one. I interact with my clients roughly once a week. Feed back and planning are important. But the rest of the time I'm deep in coding. Honestly, if I had to interface with people for 80% of my job I wouldn't do this kind of work. Having also worked for a mid-sized dev company, I can attest that those sorts of disruptions from "politics and dealing with diverse situations outside the screen" were nothing but a productivity hit for everyone. Maybe some environments accept that as normal. Doesn't make it a good thing.

Rant about disruptions aside, that was my comment (personalmis) showcased previously. Though very new to AI myself, I'm embracing it. Not because it's trendy, I held out for ages. But now that I've warmed up to it, I find it to be a great "tool" as Arjan describes, to sort information, get answers quickly, and so on. Much like the rubber duck. We all have one of those right? Got a problem, talk to the duck. Well, now that duck can talk back! It reminds me of Star Trek where engineers would "write" a new holo-deck program just by describing it. Fascinating.

[lancemarchetti]

As a beginner coder, I can honestly say that AI has helped me understand how js syntax works in basic terms and it explains what each line of code actually does. It was difficult sifting through around 50 websites to construct a comprehensive understanding of Javascript. So I'm supercexcited at the arrival of AI with regard to coding.

[malikgaruba]

Would suggest being general and use AI as support to ease development / implementation...

[marioluizbernardinelli]

Hi, Arjan. Another magnificent video, this one makes us think. I think AI can replace programmers and developers, may be no completely in a near future, but it has power to do it.

[getoutmore]

Any more follow up ideas for specializing in Cybersecurity? Can you go a bit more in-depth?

[eritech]

I feel good when I heard the last words

[ViktorFerenczi]

Those developers who are not afraid at this point must ignore the reasoning ability of LLMs. This is only the beginning, they will only get better and cheaper from this point on. The AI tooling will be there in a few years. The best we can do is trying to keep up with the tooling. Whether we'll see a massive job loss depends on how companies decide to "spend" the extra efficiency. If it goes into getting rid of technical debt and building new services, then there will be no job loss. If they want to pay that out to the shareholders, then most of us will be doomed. I guess reality will land somewhere in-between. But as always, nobody can see the future... - Just my 2 cents.

[fmaion]

you are never "just another youtuber". True be told: You >> Other YouTubers

[MrMariusvdw]

Customers would have to explain to AI what they want. We're safe.

[Deadlious]

Specilize, but don't close all the doors. Keep your mind open for new stuff. If you specilize too much in one direction and do not bother to get to know other stuff, one day you would be a dinosaur with a ton of knowledge, that just a few would need.

[ahmedshahriarsakib]

The sooner you think you can learn coding utilizing ai tools, the faster they release more advanced automation tools to replace u utilizing the same ai tools

[cetilly]

I think one of those vanity comments “accidentally” made its way into your video 😂

[aleksandristomin]

"They're never going to be able to create anything new".

Well, tell that to a digital artist. The only reason it "can't create anything new" is because art has no "borders" and programming still does, but that's only because the abstraction level of programming isn't very high yet, but it will be in the future.

I'm 100% sure that AI will replace 99% of programmers in the future, because it will be able to create new ideas and code.

[frozeneye]

Good point and tools yes. But the boiler you still need to understand the code because my boiler plate and another person boilerplate looks different.

Example Django… I spent time trying to use ORM. Pain in back side and just sql will be quicker and easier.

I did eventually get a maybe to be heavily optimized queryset using a single sql query apart from the two which is built in by Django. So it is not best way for some since everything offloaded to F. This means the DB makes the tables and calculations based on query. Not loaded into python side. This is not great choice for most people but if the Django runs on a small system when the database is still super horizontally scalable, it makes sense.

But I agree and disagree… issues are already entry level is considered 10 years experience. It is already complicated and then this. I am still learning how to use python which is considered easy… it really is becoming stupid to code I think and understand why so many people turn away from it. Back in the day and even now we had forums etc. good old linuxquestions.org still going strong. On that forum back in the day it was really people helping people and still mostly is. But it is becoming a bit like other popular platforms where people are just trying to show off crap or not help the question asked. Not complaining about how people behave but between most of it being oh you using windows instead of mac and you using Ubuntu instead of Arch or ohh you using rhel so just call their support does not help anyone. Software programming is already a massive endeavor which no can degree will prepare you for unless you combine it with electronic engineering and and and .

Just the word oh it replaced DLL… ask most people the difference between a DLL and a .so they won’t know. Even though on windows and mac, a lot of what used is actually .so because of apps installed.

Not dooms day but it really is becoming too much to learn for a lot of people unfortunately. You are better off training to be a brain surgeon or heart surgeon at this point because it is not expected to know as much about every single library or way of doing things, they deal with a physical existential problem and pay more and when retired you go teach at university. Medicine really is getting to a point where it seems better option compared to IT especially with the demographics of countries.

Just imagine a kid now trying to get things done. Super complicated compared to before. Before you could learn on the job. Now you are expected to be young and being more knowledgable than the devs that started the company have at their age.

Same issue with being a pilot. Just for what is considered entry level I do not want to do 7 interviews. I get if I apply for fang ceo position yeah sure. But it is becoming stupid to be honest.