234. What Is Next For Software Development?

[@j2csharp]

Yes, New Outlook needs a new outlook, please.

[@sadhappy8860]

Oh good lord yes!!

[@IAmTimCorey]

I want the VS Code of Outlook - something simple and quick that does the job well and is also extensible enough to fit any scenario without forcing that scenario on everyone.

[@j2csharp]

@@IAmTimCorey I just want the missing parts from the original Outlook to be implemented. It falls short.

[@timyoung6495]

I really appreciate how balanced your take on AI is.

[@IAmTimCorey]

Thanks!

[@diegomilitao1400]

Hi Tim, I know my question has not much to do with the subject of the video, but I was wondering what you would recommend in case of thinking about creating a mobile app, registering in the stores, cross platform mobile app and everything else. Every time that I search about building a mobile app, I hit in React JS and React Native, and there are so many comparisons of React Native versus Maui which React is the "winner". Could you please give me your thoughts about building a cross platform mobile app nowadays? Thank you

[@IAmTimCorey]

There are lots of options because the different options fit different scenarios better than others. So I can't say one solution is "the best" because that's not true. Comparing React Native to MAUI, for instance, isn't a great comparison. That's like asking if a bulldozer is better than a SUV. Both have a purpose and both do a great job at what they were designed to do. Both can do the job of the other one, but that's not a good idea because they would be rather poor at doing the job best suited to the other. If you are already a C# developer and have the rest of your code in C#, it makes more sense to use C# for your mobile front-ends rather than trying to implement an entirely new language (JavaScript) and framework (React Native) for your mobile UI. That could mean MAUI, but I would recommend you look at Uno and Avalonia. Both are good options for cross-platform applications in C#.

[@diegomilitao1400]

@IAmTimCorey thank you Tim for the explanation. Uno and Avalonia are new words to me, never heard these before and I'll take a look at them. Someone told me to check it out Flutter too, so I have some homework to do on this. Thank you so much for spending your time answering me Tim, you're the best!

[@faisalalhoqani6151]

Great episode thank you dear Tim and keep it up.

[@IAmTimCorey]

You are welcome.

[@user-yr1uq1qe6y]

Companies are throwing AI into everything either out of fear of missing out and falling behind (larger industry leaders) or because VC money is only going to fund because of the same fear. The problem is that it’s just being hacked on to the side of things right now.

[@IAmTimCorey]

Yep.

[@torrvic1156]

Thank you for your valuable opinion Tim! AI topic sounds really scary and it surprises me how many young developers I saw relying heavily on it. But you can see that it was written by AI when you ask person about details of their code. Honestly it scares me that it can be possible that after 50 years or so AI can become a mature technology and will make all devs obsolete.

[@IAmTimCorey]

Don't worry about that. AI won't be able to do the job of developers. It will become an increasingly important tool to help developers, but that's not the same thing as replacing developers. Besides, the global market for developers is growing, not shrinking.

[@ABMedia83]

Yep, sometimes Copilot spits out the wrong code, and if you don't know what you're doing, you're going to have problems. Don't get me wrong, it's a great help, but it won't replace the programmer.

[@brandonh5142]

That's like saying automated driving will never fully take over because sometimes you have to take over the wheel. AI will get there eventually. It'll take time but it's already making us way more efficient to the point companies can scale back hiring.

[@IAmTimCorey]

@brandonh5142 - I don't think your illustration supports the point you are trying to make. Let's look at self-driving for a minute. Let's start by defining what companies have to do to get self-driving to work. They have to operate a vehicle with a limited number of options (gas, break, turn - all in 2 dimensions only) on a closed system (only operating in vehicle-specified areas like roads and parking lots) with a well-defined set of rules (traffic laws). Yet with all of these simplicities, companies still don't have true self-driving. Sure, Waymo has vehicles driving without drivers in the vehicle. However, the remote drivers have to take over to fix problems. Plus, city driving is much less complicated than rural driving, where roads don't have clearly defined lines, where construction makes a mess of things, and where snow, mud, ice, and dirt all conspire to make driving difficult. Software development is a LOT harder than self-driving. We don't have a limited number of things we can do, we don't have a well-defined area to work in, and we don't have well-defined rules to work inside. Otherwise, we wouldn't need to write code at all. Think about it - all the code we ever need has already been written. We should be able to just put it together like LEGO pieces. But that's not how the real world works. The real world is messy. It is analog. There was a person who documented the steps that were taken to determine if a user should be notified about a Tweet. I don't have the link on hand, but there was something like 50 steps that had to be taken for that one action. That wasn't because of inefficiencies, that was because life is complicated. That's not something an AI can do well. So no, AI isn't going to "get there eventually". It will get better, but it will not become able to somehow "self-drive" as a software developer. That's just not possible.

[@andywalter7426]

Do you think ai can be great for when learning about c# as far as coming up with practice projects? i asked ai a question and it suggested if a person uses a paid account that uses gpt 4.0 instead of 3.5, then it can be smart for coming up with practice exercises for different areas of c#. does that sound like a good idea. my biggest struggle is coming up with practice projects. your course was great but only had some (not even more of them).

[@hqcart1]

Tim, why aren't you talking about AI IDEs? it's a big topic...

[@VeaceslavBARBARII]

Because an AI IDE lowers the barrier to entry into programming and boosts productivity, but does not take away the complexity of software engineering.

[@valhallagalex]

The biggest challenges for you are architecture related. Any decent programmer will remember the syntax. AI is a glorified auto complete.

[@flogginga_dead_horse4022]

@@valhallagalex AI is a glorified auto complete. That's hilarious

[@AlecMaly]

He did: "Don't let it drive the bus"

[@hqcart1]

@@AlecMaly IDEs drive the bus dude

[@p199a]

Tim you should try out CS50 AI (its harvard computer science course that has its own AI that help you learn the course) its amzing. Its spot on 100% of the time. It explain every topic perfectly with perfect examples. and its free

[@IAmTimCorey]

I'm glad you enjoyed it. The only way an AI can be spot on 100% of the time is if you were working with a simplistic training set and easy, well-defined problems. That's rather common for intro-level courses. The issue is that this isn't the type of code you typically work with in the real world. Real-world code is complex, messy, and not nearly as clearly defined as you would get in school. That's when the AI starts to hallucinate more and answers become less reliable.

[@andres.biarge]

Imagine the discipline I have to exercise being a Power Platform developer: copilots everywhere!

[@IAmTimCorey]

For sure!

[@ian2neko]

Regarding cross platform web app, which one has brighter future and larger community (finding support)? PWA? Electron? .Net MAUI? Tauri? Or something else? Thanks

[@IAmTimCorey]

PWA already has a very large footprint because it is a web standard. It isn't something directed by one company/organization/team. Beyond that, these things will change and shift over the years.

[@sadhappy8860]

Thank you for saying AI isn't magic. I do use it, it is a tool after all. Recently though it gave a response to a co-worker about a library and I had to say that the library is now deprecated, luckily before people went off coding. They then asked the AI and it was indeed deprecated. Use it by all means but be skeptical and double check what it says!

[@IAmTimCorey]

Absolutely.

[@lolyasuo1235]

AI is just a trend

[@asdf8948]

Yeah 70 years of AI research is just a trend.

[@sadhappy8860]

It's not a trend, it is another tool for the toolbox. We have to live with it now.

[@lolyasuo1235]

@@sadhappy8860 It makes you write worse code if you rely on it.

[@IAmTimCorey]

The idea of forcing AI into everything is a trend, but it is one that probably won't go away anytime soon. AI is here to stay. However, the over-use of AI is something that will be reduced at some point.

[@andywalter7426]

I like the suggestion you have about learning about ai. the only problem is none of your courses even talk about ai. i don't even know where to find the information to figure out the formulas about when to use ai and when not to use ai. i had cases where i used ai for a part of programming where i did not understand. in those cases, if the answer was not correct (because the end result was not what was expected like parsing csproj files), it would recalculate and eventually get the correct results. the truth is its very common to use code a person can't understand but trust it works. even open source is like that since you don't really know why it works like it does when it works. if a person had to understand every little thing it would do, then the best solution is to just invent everything yourself.

[@IAmTimCorey]

A lot of this is about experience. You don't want to use AI when you are learning C#. You want to learn C# well. Not just the syntax but actually gaining deep experience in building real-world applications in C#. Then, it will be MUCH easier to use AI and understand how and when to use it.

[@andywalter7426]

@@IAmTimCorey i actually know c# very well. so far, it was obvious when it was giving me incorrect answers and i implemented my own solutions. it was great for cases where i needed help with naming or how to organize the classes (a start on interfaces). was also great for refactorings or if i knew the code but knew it could do it much faster like hard coding a list of information for test purposes or if i have code in one format like getting from iconfiguration and need to change to getting from text file when its very routine).