शैडोइंग अभ्यास: 5.4k - वीडियो के साथ अंग्रेजी बोलना सीखें

A lot of you have asked me if I think that AI will replace software engineers,

and I want to talk about that and share my opinion in this video.

So first of all, I basically use AI to write all of my code now.

Like 99% of my code is written using AI.

And I try to use the latest models and tools as they are released.

So right now I'm using the model Claude 4.5 Opus,

but it may be something different depending on when you're watching this video.

So I try to use these tools as much as I can,

and as I use these every day,

I'm not seeing how they are replacing engineers.

The way that I see things is that there are two main ways that I see myself and other people using AI.

The first way is called vibe coding,

when we're telling the AI to work on an app or a feature,

and we don't look at the code at all.

We let the AI handle all of the code itself.

And the other way that I see people using AI is just regular software engineering.

So like we tell the AI to generate the code,

but then we also review the code to make sure that we understand exactly what's going on.

And by the way, I do both of these things.

So I have a side project that I like to work on

and I use vibe coding there because I just want the product to work.

I don't really care about how the code looks.

I just want something working so that I can get users and I can get some feedback.

Now, when I'm working at a tech company,

I have to use regular software engineering because,

for one, I need to understand exactly how that code works.

That's part of the responsibilities that I have.

And number two, I've also noticed that vibe coding or letting the AI handle all the code actually does have its limits.

But before I talk about that,

in using AI to vibe code and to do regular software engineering,

I've noticed that it's a very collaborative process,

both when I'm using it and when I see other people using it online and in real life.

So we're basically giving the AI a task to do.

The AI comes back and then we check its work to see if it's correct.

Maybe we need to fix something.

Maybe we forgot to tell the AI something or maybe the AI misunderstood us.

But in all cases, we're going back and forth with the AI

and giving it tasks and then adjusting based on the results and the work that it does.

And also sometimes maybe I don't understand something,

so I would ask questions to the AI and we would have a whole conversation back and forth.

So basically, as I'm using these tools,

I'm thinking to myself, it's a lot of back and forth collaboration with this tool

rather than something that is replacing me or replacing someone who's using this tool.

And so I don't really see the angle of AI replacing software engineers

or AI replacing anyone that's using this tool because it's a very involved process.

And I think that to get the most value out of it,

it does have to be an involved process

because you want to give the AI the most detailed instructions and the most detailed direction that you can.

So it does involve a lot of collaborating.

Now, I want to go back to the point that I talked about

where I've started to hit some limits with vibe coding.

And those limits come when your app gets a lot more complex.

So basically when you're vibe coding and you don't look at the code at all,

the way that you would check the AI's work is you would start up your application

and then you would click around and use your app to see if it's doing what you wanted it to do.

If there's some mistakes, you go back to the AI and tell it,

here's the problem and try to fix it.

And then you check again and you don't look at the code and you let the AI manage that,

but you're going back and forth in this way without looking at the code.

Now the problem that I see with this is that when I started working on more complex applications,

like production level applications, a lot of times it's actually not easy to start up the application and just try it out.

A lot of these applications are dependent on different services,

and I often have to get my account into a very specific state to run my code.

So that involves generating a lot of data to support that account,

or sometimes changing some if statements to be true all the time,

just so that the code goes through the flow that I'm working on.

And also, just imagine you're working with five other people,

and everyone's adding to the project at the same time.

It becomes really difficult to know what has been added,

what has been changed or removed without looking at the code or working with the code.

And what I found is that in these situations,

when the code gets complex enough,

it's actually easier to just look at the code and work on the level of the code.

So instead of working on the app,

I'm actually just working on code.

I check if the code that the AI generated is correct.

Is it maintainable?

Do I need to reorganize it or refactor it?

But in this case, the code is actually the final product.

And I work on the level of the code because I found that in many situations,

working directly with the code is actually the most efficient way to get things done and to check the AI's work.

I think that a lot of people have a misconception that code is like an obstacle that we need to solve

in order to get what we want to get done.

Like in order to create the feature,

we need to figure out what the code is for that feature.

But what I've noticed in working with AI a lot on these production apps is

that a lot of times the code itself is the value

because code tells you exactly what the software is going to do.

So if I want to understand something very deeply,

I would just look at the code directly.

It's precise.

Basically, it tells you exactly how things are going to work.

It's not like a summary from the AI that may or may not match up exactly to what you expect.

And it's also concise.

So sometimes maybe the AI over explains things and there's too much content, like it's overwhelming.

But I can look at the code and I can tell exactly

if the AI has handled all of the requirements that I gave it,

or if it missed something,

or if it misunderstood one of my requirements in my prompt,

100

and I need to give it some corrections.

101

So basically what this is saying is that coding is just another skill

102

that you can have and that you can reach for in certain situations.

103

And we use it in these situations mainly because it's more efficient than just using the AI.

104

And I don't really understand why people are saying that you don't need to learn coding anymore,

105

or that you can just learn AI.

106

Because if you learn coding and you learn AI,

107

wouldn't that be better than just learning one skill?

108

So I've often heard the phrase that AI is not going to replace you,

109

but someone using AI is going to replace you.

110

Now, I haven't really seen that happen in software engineering

111

because all the software engineers who have the coding skills and who have all the other core software engineering skills,

112

they're starting to learn how to use AI and integrate it into their workflow.

113

So you have all these people who know software engineering and they know AI.

114

So I don't see how someone who just learns AI

115

and avoids learning to code is going to be able to compete with people who know both of these skills.

116

Now another idea that I've heard is that this is the worst that AI will be right now.

117

It's going to get better in the future.

118

Now, will it get better?

119

I'm sure it will get a lot better in the future.

120

How much it will get better by?

121

I don't think that I have the authority or the qualifications to be able to predict this stuff.

122

But what I will say is

123

that I feel like the people who are researching AI are a little more qualified to kind of make these predictions.

124

And what I've noticed is that these AI researchers,

125

a lot of them are hyping up AI a lot on X and on other social media.

126

But when they actually release these models,

127

I feel like the models are incrementally getting better,

128

but they don't really match the hype that they're giving out.

129

So that's just my observations there.

130

The models are getting better and better,

131

of course, but from using them every day,

132

it's not like a incredible rate of progress that it's going to replace people or replace engineers soon.

133

The other argument that I've heard is that the current AI tools will make engineers more efficient.

134

And because engineers are more efficient,

135

we're going to need less of them.

136

So it's basically the same as replacing engineers.

137

Now, I kind of understand the logic of how you would think about this,

138

but the thing is that a lot of things make engineers more efficient.

139

I think that tech companies have a choice.

140

They can use the same amount of workers to create more products and more projects,

141

or they can create the same amount of projects, but with less people.

142

And I think that for the most part,

143

tech companies tend to keep the same amount of people and try to work on more products.

144

Tech companies are rewarded for growth.

145

They're not rewarded for making the same amount of money every year,

146

they're rewarded for making more money this year than they did last year.

147

So there is a financial incentive to,

148

instead of maintaining the number of projects they have with less people,

149

to have the same number of people,

150

but have more projects and more products in order to grow their revenue.

151

However, in saying this,

152

I don't want to create the feeling that it's easy to get a job

153

because I know that the job market is really tough right now,

154

and I'll probably talk about that in a different video.

155

But what I do want to point out for this video is

156

that all that you see on the news are the layoffs.

157

That's all they're going to report.

158

What you don't hear about is that a lot of the companies that did the layoffs in the past,

159

they've been rehiring for the past couple years.

160

Remember when Google and Meta laid off thousands of people a couple years ago?

161

Well, Meta has been rehiring thousands of people over the past few years,

162

and so has Google.

163

Google is actually close to their all-time high in their employee count.

164

Now, this doesn't mean they will not do layoffs again.

165

They might do more layoffs,

166

They might rehire again and do more layoffs and rehire.

167

But the point I want to make is that in the news,

168

you're only going to hear about the layoffs.

169

If companies decide to hire again,

170

you don't really hear any news about that.

171

And it creates this impression that jobs are just going down all the time and they're not going up again.

172

And, you know, that kind of perception,

173

I think, is not entirely true.

174

Even though the job market is very tough right now.

175

The perception that this industry is collapsing is not true at all.

176

And if you look at the actual employee numbers for many of these companies,

177

you realize that there are way more people

178

that are making a living in tech

179

and in these tech companies than there were back in 2018 or 2019 when we had a good and strong job market.

180

So I think that the story about the job market is not so one-sided and negative.

181

I think it's somewhere in between,

182

and it's not an indication or evidence that AI is replacing software engineers.

183

In using AI every day,

184

I don't see this as a tool or a process that is replacing engineers.

185

It's very collaborative, and it requires a lot of input from the person using the tool,

186

and in many situations, coding is actually the most efficient way to do things.

187

So all the core software engineering skills that you have are actually very valuable.

188

And that's why someone

189

that just knows AI is not going to be able to compete with someone who knows engineering

190

and they know how to use AI effectively.

191

So if I was to say,

192

do you need to learn how to code?

193

I would say, yes, you definitely need to learn how to code

194

if you want to work at a tech company as a software engineer.

195

And I would phrase it like this,

196

learn AI and learn how to code,

197

not learn AI instead of learning how to code.

198

Because by learning both of these skills,

199

as well as all the other software engineering skills,

200

it'll help you compete with everyone else who wants to work in this industry.

201

So thanks so much for watching.

202

I hope this was helpful for you.

203

Now let me know if you have any questions or thoughts.

204

Obviously, I don't know everything,

205

so we can have a discussion.

206

and let me know if there's other topics that you want to discuss in future videos.

207

Thanks so much for watching and I'll see you in the next one.

शैडोइंग तकनीक क्या है?

शैडोइंग (Shadowing) एक विज्ञान-समर्थित भाषा सीखने की तकनीक है जो मूल रूप से पेशेवर दुभाषिया प्रशिक्षण के लिए विकसित की गई थी। विधि सरल लेकिन शक्तिशाली है: आप मूल अंग्रेज़ी ऑडियो सुनते हैं और तुरंत इसे ज़ोर से दोहराते हैं — जैसे वक्ता की छाया 1-2 सेकंड की देरी से। शोध से पता चलता है कि यह उच्चारण सटीकता, स्वर, लय, जुड़ी हुई ध्वनियाँ, सुनने की समझ और बोलने की प्रवाहशीलता में काफ़ी सुधार करता है।

शैडोइंग अभ्यास: 5.4k - वीडियो के साथ अंग्रेजी बोलना सीखें

संदर्भ & पृष्ठभूमि

दैनिक संवाद के लिए शीर्ष 5 वाक्यांश

चरण-दर-चरण शैडोइंग गाइड

शैडोइंग तकनीक क्या है?