Best and fastest way to grow along with AI?
We all are using AI a lot to build things but how do I make sure that I don't become handicapped due to it. If someday it stops working I shouldn't be completely dependable on it. Even though I do understand the basic foundation and concepts, I just feel for me AI just saves time. The time I used to spent looking on stack overflow and GitHub repos i can get the best working code with AI.
But the question is with all this how do I make sure i learn 10x faster now. As I have the best code for that feature already available to me.
One interview, 1000+ job opportunities
Take a 10-min AI interview to qualify for numerous real jobs auto-matched to your profile 🔑
AI will give whatever code it can - some of it will be correct, & some of it will be buggy with hallucinations.
But that doesn't do anything to change your human brain's cognitive & learning capacity. So you will still learn at your own 1x human speed. That won't become 10x just because AI is hallucinating some bug-filled code for you 😄
Yes. This.
I use the Cursor pro plan. Even with their latest best models, I still sometimes find it either overdoing what I asked or unnecessarily wasting my tokens by doing absolutely random shit.
For ex, yesterday, I was working on some NextJS SSR related issues. It did correctly identify that my hosting service is not configured to connect to my secured DB as my DB was not publicly exposed and was locally hosted, instead of telling me that I should provide a proxy access to my DB for my NextJS app, it suggested that I refactor my app from SSR to CSR. I was like "huh, okay, let me see what shit its gonna do". And then it spent a good 10 minutes refactoring my entire codebase, and it still failed to migrate my app from SSR to CSR.
Inspecting its thought process, it correctly identified the solution that it should replace my SSR methods with useEffect or the external store sync hook, but it ended up doing everything else like memoization, adding more CSS (god knows why), breaking an already small component to even smaller component unnecessarily in the name of reusability while it was only used once in the entire freaking codebase.
If you look around, every f***ing AI model and code assistant is deeply trained on the web stack the most. All demos show the same. That they build fullstack web apps and everything with just a single prompt. But in reality, it can't maybe, max to max, replace a junior engineer, which I still wouldn't recommend because few years down the line, when I move up or move out of coding tasks towards team management and office politics games, there's gotta be someone with good fundamentals who can use AI aid to get things done.
You just gotta accept the fact that you are a biological human and you have physics limits to how fast you can learn something. There's no 10x faster, when an AI model can learn 1000x faster with additional hardware and memory and better algorithms hooked to it.
You gotta focus on the fundamentals IMO. The absolute basic building blocks.
If you are solid with that, move on to design and architectural knowledge. Learn patterns, read papers, explore codebases on GitHub of projects that are interested to see how they do what they did that caught your interest.
If you are solid with that as well, you'll now have to think like a product manager. Know the people who use what you build as a dev, interact and learn from them as to what can be improved, what they don't need or find pointless, what they actually need to aid them in better ways. That's what you should be doing. This is the ultimate step. Because, the end user of all products is always a human with a bunch of bots and automation in between the end user and your product. Do what sells. Do what brings more satisfaction and smile on people who use it. Do what brings in more revenue to build even better things at scale.
I'm no super senior dev. I'm just an SDE 2/3 from a corporate world POV, but these are the lessons I've learnt from people at the top when I interacted with them, attended their sessions, conferences, reading their posts and news letters etc.
I dont beleive that AI can make things superfast and you can just use it blindly without fundamental understanding of your work.
If you are working on a small scripts or codebases...you will have to do 3-4 iterations before you actually get something working that you want but there are still chances that it may not cover many corner cases. Which means you have to either understand the code and make changes manually or use more iterations to modify it which may also have some issues. Code trustability and verification is big issue with AI generated code. Which means you spend same or equal amount of time to get the same work done with AI for small codebases.
If it is a large codebase, then the problems are even bigger. There are more "WHAT IFS" than "AI CAN DO THESE SUPERFAST"
Or do we blindly trust and integrate whatever AI produces for getting the given work done.....which may create 100 new problems in future for some different input? How much extra time will we spend to solve them?
Not even considering how optimized the AI generated code will be where the requirements are so strict.
There is no way people can compete it AGI. It's just a few years away. It will exceed IQ level of all human beings.
We all are already handicapped. There's no going back. Simple. Accept the reality and live with it
Use AI to get the blueprint, validate the blueprint probably ask skeleton code.
Using AI to generate entire code can be sloppy and code which is not required in the first place. Most people are not good at giving the expected prompts and the required details.
Many devs have become very lazy now. Just dependent on AI and code reviewers have a bad time reviewing their code.
Test cases should have proper use cases like P0-P2, instead these devs ask AI to cover 100% of lines and say its done. Many test cases are useless and pointless.
Upper management doesn't care they show that what ever module they own are 100% covered.
When something breaks in production it a one of case.
When computers or calculators came we stopped doing manual time consuming things and delegated those things to machines. But look where we are now, building better and bigger things.
AI is just another tool to help us to build some even better and complex things that were not thought of till now.
Use AI to learn and improve rather than doing the job. Same was with stack overflow, brainless copy paste will surely make us obsolete.