# AI's Cognitive Shift: Judgment, Workflow Redesign, and Engineering Discipline

**Podcast:** Tech Lead Journal
**Published:** 2026-07-13

## Transcript

What you are getting paid and what you pay for is always a scale.
You'll never pay for typing speed.
Today's guest is AJ Gore, former group CTO of Gojek and operating partner at Peak15 Partners, who scaled Gojek from 300,000 to 120 million monthly orders.
Today, he builds and advises at the intersection of AI, infrastructure and product.
AI brings a lot of anxiety to people today.
We think we're all going to have no job in the future.
But what does this AI era give you perspective?
The first perspective is engineering practices are getting reinforced.
thing, these agents will reward us as builders.
We can build things at different speeds.
The problem is people are trying to use AI at a very minimal level.
They're not trying to offload their job and think more.
When expertise is becoming almost free, whether design, security, audit, everywhere, then what becomes expensive?
And my answer is judgment becomes expensive.
And judgment layer is going to be very important for everybody.
Because if agents can do good things fast, they'll do bad things fast as well.
For leaders out there, many are thinking AI means I can reduce a lot of people's layoffs.
What is the risk?
It's not about people reduction.
It's about basic job going on.
Structured work is getting at least.
Cognitive work is getting at least.
If they're reducing it, fine, but they're creating cognitive gap.
They're not going to go faster.
They're going to go same speed.
If they're okay with that speed, let people go.
It's fine.
But if you want to accelerate something, then instead of letting people go, think through and enable them with more skills.
88% of AI companies are out of AI.
Corruption has increased, but only 6% have successful projects.
So why?
Hey, quick pause.
My goal with Techly Journal is simple.
Learn from the best in tech so we can all grow together.
If this resonates with you, Hit subscribe to follow the channel.
It's the biggest way for you to support the show and help us keep bringing great guests and insights to you.
Thanks for being here and let's get back to it.
Hello, welcome to another new episode of the in-person podcast recording of Techly Journal.
Today, I'm very excited to meet you, Ajay.
So if you remember, five years plus ago, we did an online recording, I think episode 23.
Very excited to have this chance to actually talk to you in person.
Welcome.
Yeah, thank you, thank you so much.
That was like COVID, I think, if I remember.
And we were logged in our rooms and we were like kind of recording and trying to do things.
This is much, much better.
It has always been exciting to actually revisit what we have done in the last five years and see the perspective change.
And I think COVID changed many things.
One of the things, if AI has accelerated the coding, COVID has accelerated the digitization.
Yeah, and also perspectives of life, I guess.
Yes, yes, yes.
And COVID also created some false hope, saying people who work remote can work remote all the time.
And then after some time, people realize, no, no, no, no, that's not true.
We need people in person.
So we are social animals.
We need interactions.
We need ideas to go through and all this stuff.
So I think there are also lessons in terms of what you can digitize and what you cannot digitize.
But that...
evolution would have happened over many, many years, but COVID just forced that acceleration so fast.
Yeah.
And yeah, hence we have this in-person podcast recording.
So I think that's one part that I feel digitalization does not give you the same kind of like interactions, the richness.
Yeah, it always is like very different in person.
Yeah.
Also what happens is, I'll tell you this, when I was meeting so many people on Zoom, So you only see still this.
So you don't know what the person really looks like.
And then sometimes what happens, you assume that person must be very tall, and they're not that tall.
And sometimes you assume that person won't be that tall, but they're very tall, or whatever way.
So you get always surprised, oh, I did not think like that.
Has that happened to you at all?
Yeah, for sure, for sure.
Okay, cool.
So I'm not the only one who does that.
Yeah.
So I think...
I've seen your posts, your blogs, and all that.
There's a little bit of change in terms of life, right?
I think maybe like three months plus ago, you stopped your last role in TIG15.
So tell us maybe what have you been up to?
What's keep you busy this day?
So I wrote in that blog post, sometimes you run so much, or you walk so much, you talk so much, or you work so much.
you assume, you accept that is a new normal.
Correct?
And then you don't know when to stop, what to do.
You feel, you don't understand, oh I have been running all the time.
What it feels like to walk?
What that slow pace means?
What is a slow thinking?
What is being slow and having time means, right?
So, and there are multiple...
There are multiple reasons I did that.
For example, there are some things around family which I want to focus on.
There are some things around myself which I want to focus on.
And one of the thoughts which also came across is like we work a lot.
All of us earn money.
We work for many, many years and we spend money.
But we forget that we earn time as well with that money.
And we end up spending time most probably at the end of...
When somebody calls retirement, I don't believe in retirement.
But yeah, I'll do that.
I'll chill at 65.
Yeah, but that is something like we are doing it when you have money and time, but you don't have health.
So I wanted to bring that perspective, and I'm fortunate that I can do that.
It's very difficult to do.
I can tell you that.
It's very difficult to do.
There are some external forces.
There are some internal forces.
But sometimes it just feels like, let's do it.
Let's pause.
Let's do some tinkering.
Let's do something else.
play those things which you have been willing and wishing to play for many, many, many days.
So that's what I'm doing now.
I am going after my passion.
I am building things.
And then I feel very much associated with the youth mental health and other issues.
So I am volunteering with the Hello Ally or Ally.
Ally is an Indian not-for-profit which is building mental health solutions to train therapists, teachers, and related people to identify and help youth.
And that's very good for us.
And everybody should actually do something like that.
So I'm doing that.
I'm doing that.
That is my volunteer work.
And then I'm helping few organizations.
I just accepted being an advisor to NUS Enterprise because I was excited by their mission of doing things around data and AI.
So I'm doing that.
And then on the other hand, we are just going through this craze of bots and this and that.
And I have been tinkering around that.
We built Cloy Station just for sake of it.
It is OK.
It's side hustle.
If it takes off, it doesn't take off.
I'll go build, tinker something else.
So yes, and reading book.
And as you know, when you call me saying, Ajay, can we do this?
And I go, yeah, I can do this.
I have time.
I'm spending time.
So I'm spending time with friends like you.
So yeah, that's what is going on.
It's like pretty exciting journey in that sense because what is happening is that you create slack in your life and then slack allows you to think through.
When you meet people without any agenda, then you have different flow of conversations.
They go in many directions and then you come out mostly happy.
Oh, I felt good.
meeting this person.
And then we talk about, because we don't have any agenda, we don't have any transactions to execute, we don't have anything to go after or before that.
Then the fluidity of conversation is very positive.
So that is what is happening.
And then a lot of people are reaching out to me, asking me for advice.
They're in mid-career stuff.
They are going through this AI stuff and they're scared.
Some of them are excited.
So I'm telling them what we should not do because I can't tell you what to do, but I can tell you what not to do from my experience, which actually kind of reduces the possibility of going and doing something stupid.
Yeah.
So I think you brought a very interesting perspective, right?
Because if I look at your career, you've been kind of like hustling, or let's call it hustling, like working hard, you know, from...
From ThoughtWorks, you know, spend 10 years, Go-Jek, maybe five years.
Sequoia, pick 15, it's about another five years.
So you've been like kind of like running in full mode all these years.
And I think when you mentioned about earning time, spending time, I think it's in everyone's interest now to actually understand what does the concept mean.
So maybe if you can elaborate for people, especially mid-career and, you know, later stage of career people.
Actually, sometimes we don't realize that actually we miss this opportunity to create slack time, spend time on our passion and all that.
Yeah, so I wrote this blog on my blog, the space between chapters.
So I'm living in the space between the chapters.
One chapter is over and the next chapter is supposed to start at some point of time.
And I found this space, or I kind of created this space.
And I'm going to live through that space.
A lot of time.
What happens is people don't believe that they can create that space.
That is first thing.
If you don't believe that I can create that space, then you will not be able to do it.
And I started that blog saying the most courageous thing is to do stop running.
It's like...
Just stop running.
That's the most courageous thing to do because there are a lot of things which we have preconceived notion.
We worry about a lot of things which are not yet come, and we ponder and regret about a lot of things which have gone in the past.
It's so difficult to implement that whatever is happening in the present, can you enjoy that?
And I'll tell you, Henry, last three months, it's like February, March, April, it's gone.
We are sitting on the 30th April over here.
And it's so funny that three months and 90 days are gone and I'm like, I'm still like busy and meeting people and I'm enjoying it.
I don't know where those three months are gone.
So first thing is that I spent a lot of time with my friends and family.
I took my kids to bouldering, which I never thought I would do, right?
They climbed the thing and all that stuff, which is amazing.
And then I'm again going to spend a week with them just like that.
And then what happens is the slack allows me to think and that is very important.
So I will try to give you the analogy of getting up late and getting early.
So if you get up late in the morning, what happens is suppose you have to go to office at 8.39 and you get up at 8.15, then what are you doing?
The first thing is you are reacting to the time.
So you are hurry, hurry, hurry and then you...
Somehow you get into MRT or car or whatever and get to your office by like 9.
And then whole day is going on reactions, reactions, reactions, reactions, reactions, and end of day you come home and then you actually just do default things.
But now just flip this and say I get up at 5.
Now I can think and deliberate on things and I can create a slack.
Yes, I have to sleep early at 10, but that slack is way more important than actually working like a night owl till 2 o'clock and then getting at 8.15.
Actually, we don't work.
We actually spend time during work hours, but not usefully spend time.
I'm not saying that don't go on social media.
I'm not saying that don't watch movies.
I want to watch movies.
But what I'm trying to say is create a slack in the system.
And it is very true for creating the slack, not in your personal life, but within your teams as well.
Else innovation will not happen.
A lot of engineering teams I see, they always react.
Now, second thing, many people will say, Ajay, you are saying it because you can create slack and you have money and you are financially secure.
I don't think so.
There are many, many examples and I have seen many people who don't have a lot of money, but they have done this as well.
I think I watched TEDx talk.
I don't remember the talk, but I watched TEDx talk.
Somebody was in Thailand, in Bangkok.
and he actually, I think the name was Danai, and he actually went back to his village.
And now he runs a very big, large village where he's like, creating a house is very simple.
You just dig up the soil and build a mud wall, and you have a house, right?
You don't need to do that.
But if you have to live in Bangkok, he has to spend so much money, or Hong Kong, I don't remember.
The point is, I was at talk around 10 years back, and I was like, this is amazing.
So a lot of time we have this preconceived notion of about our work and job security and all the stuff.
And that's where I go to my principle.
When I talked to you last time, I think I talked about it.
Being uncomfortable is one of the most important things for anybody.
If you be uncomfortable in your job, then you will be able to take new challenges.
Second, being replaceable is also one of the most important things in our job because if you can be replaceable, then you can do something new and you can be uncomfortable.
And if you be uncomfortable, you will acquire more skills.
That will eventually lead you to take you out of one important fear.
Will I get next job or will I have something to do or not?
If you have done enough many things, then you will do something, right?
And that is very important.
The way...
in continuous delivery or continuous integration.
The only way we can do that, like one click deployment, is because we have test harness that we know our code will not fail and if it fails, we'll know about it before we can click the button.
The button will not get enabled, right?
That is the harness which gives us confidence that our code will work.
This is also the harness which will give us the confidence that we'll get a job.
I can do this today because somewhere I have a little bit of confidence that I can get a job.
Somewhere I have confidence that I'm useful for somebody, right?
For various reasons, various experiences and all that stuff.
People do value experience.
You must have heard this joke where somebody was called for some electric or some computer error and he came and, like, I think took a hammer and like...
put it on some right place and the thing will start working.
And somebody asks like, so how much should we pay you for?
It's like $1,000.
And like, why $1,000?
Like, no, no, no, no.
Striking with hammer is like $1.
The rest is to know where to strike.
Right?
My point is people should experience, people should gather those experience.
And if they gather those experience, they'll become valuable.
If they become valuable, then they'll be useful.
If they are useful, then they can create slack and they can create a space between the chapters.
Yeah.
So I think the importance of creating Slack, you know, sometimes we see it in the literature, productivity and creativity and all that.
Definitely creating this Slack is very important, I feel.
And sometimes it doesn't have to be not having a job, right?
You can create Slack without, for example, like maybe go for running for a long time or, you know, you meditate or, you know, just do something that doesn't really fully engage you in terms of, you know, work or career and all that.
Sometimes that could let your mind wander.
Oh, I'll tell you this.
One other hack I have is go for walks, runs, but without any additional stuff with you, not even with your phone.
Now you're forced.
But another problem is like a lot of times people want to record this on Strava.
I don't know why.
I do.
I do that.
I know.
So my point is like, yeah, it's okay.
Let it go for one day and just go for a walk.
Then you'll suddenly realize there's so much of a work.
Yeah.
I go out.
for a walk without my phone and I come back in like one hour or two hour and it's like amazing.
Like I can stop at some hawker center on the way and eat some food, have some coffee or I can do something else and come back.
And you're absolutely right.
People can create slack in their life just like that.
Very simple.
See, when I was in Gojek, somebody asked me, how did you scale Gojek?
I go, okay.
The only way we could scale Gojek was discipline.
Discipline brings the scale and actually scale also brings a discipline.
If you're a scale, you can't be indisciplined.
And if you're at a scale, then you have to be disciplined.
So my point is just be disciplined.
Say, I will do this once a week, and that's okay to be disconnected for once a week.
If you can't disconnect for once a week for two hours, then there's no problem.
But I agree with you.
Yeah.
Like for me, I've been doing a lot of running as well lately, like maybe three times a week, like an hour more than that sometimes.
I think it gives a lot of benefit, especially sometimes it's just creativity flowing.
Because the interconnectedness with things that you read, maybe things that you didn't think about, somehow it blends during the session.
And you think about many things which you wouldn't think about if you are engaged with your devices.
Yeah.
So one thing that you mentioned about earning time, spending time, I think that's also very important, right?
Because we all feel that we have to work hard during young.
This is probably also true for Asian, right?
Our teaching is that, yeah, when you're young, work hard, earn a lot, save it.
And then when you retire, you kind of like spend that.
I think that is like a big misconception in life.
Too far Asians, Indians, and everybody else as well.
I have a saying saying, we live poor, die rich.
Yeah.
So you are right.
So there is a, I got inspired by this guy called, again, I forget his name all the time.
He was Semcorp.
in Brazil.
He wrote the book Maverick.
Why I forget his name every time?
He has a very amazing tech talk.
But anyway.
So he says something like this.
Think about it.
And I think I asked you this question.
And I actually wrote about it.
I wrote a small utility on my blog called life.html.
If you go to my blog at jagore.n slash life.html, it'll ask you a few questions.
And it'll show you how much life is left.
Wow.
Okay.
I really wanted to create a gadget around it, like a digital panel which beeps today's ear and tells you how much is left.
I talk about a lot of people that they live as if they will never die, but then they die as if they never lived.
And this philosophical, but it has been there with me for almost like 10, 12 years.
So there is a guy called Ricardo Semler.
Ricardo Semler has a very good TED talk.
He talks about many, many, many things.
But one of the things he talks about is, think about this, Ricardo, one day your doctor tells you that, here is your scan, you have one year to live or six months to live.
What will you do in that six months or one year?
I think your perspective changes.
And it changes very dramatically or drastically up.
up to you how you want to think through it, but dramatically or drastically, it changes, right?
And then you will do many, many, many things which you had your bucket list or non-bucket list or un-bucket list, whatever, you un-bucket many things.
You had the bucket list and you un-bucket that.
So what he said, he's like, oh, instead of waiting for that moment, I brought my terminal days forward.
So every Wednesday I go out.
with my wife every Sunday night or every Friday I go out in the afternoon for a movie.
I have actually done that once.
We went out for a movie on Monday afternoon.
It was amazing.
There was no rush.
Everybody was chill.
The hall was for ourselves.
We enjoyed.
We had food.
There was no rush over there.
Good restaurants.
Try to do that on Friday, Sunday to Sunday.
Why?
So that is where the slack comes in.
And it's very fulfilling.
So, a lot of us also, and I used to tell this to my Gojek colleagues, take leave, take off, go somewhere, do something.
But spend your leaves.
There's no pride in carrying forward your, oh, I have like 45 leaves.
What will you do with those?
People don't take leave.
They want to work.
And I'm saying, I don't see...
things as a work-life balance.
I don't see things as 0% or 100%.
No, it's not.
It's a blend.
It's somewhere in the shade of gray.
It's not like I work 8 hours and I switch off.
No, you can't do that.
If you have to work 16 hours, work 16 hours.
If you have worked 3 days night outs, I have done that, do it.
But then, slack hard as well.
And if you do that, if you practice that, then nobody's going to question you.
Then, hey, why are you not coming on Friday?
or why you're not coming on Monday.
Because they know that if there is something on Friday night or Saturday night and that required Henry's or Ajay's attention, they will be there.
So we have to build that trust and then we can slack during the work as well.
But having this kind of slack, what I'm doing right now with advising companies, doing small things, I'm having small meals.
I'm having small bites during the whole day.
But I'm still doing this.
I'm still working.
It's not like I'm not working.
I'm still working with NUS.
I'm working with a few companies.
I'm doing this.
But not in a traditional way.
I have a sprint, so I run 100 meters.
Then I take a rest of one hour.
What I learned as well from some people, yeah, the concept of work-life balance sometimes is not appropriate, right?
It's more like seasonal thing, right, where sometimes you can work hard.
But just don't forget that you also have to go through the winter season.
Yes.
Maybe take a break, chill a little bit, not running 100% all the time, which probably is...
I have done that during my Gojek days, and I was very proud of that.
And I realized I was like, shit.
I wrote about that as well.
I talked to a blog post called Pit Stop.
If you don't take a pit stop in your car and just go around, the tires will bust.
So yeah, they go fast, but they take a pit stop as well.
Yeah, and I like the, like you brought the importance about life, right?
So I also follow stoicism and all that.
They also have the same concept, like memento mori, you know, love for the love of death.
So, you know, the illustration of number of days left, and then you just cross it every day.
I think it gives you perspective that actually, yeah, life is kind of like limited, right?
So if you don't realize how much left, you wouldn't be able to spend the time.
And also it happens, like you ask this, what for?
And most probably, so people, who have been extremely successful or extremely not successful, they reach a peak and either they go down in the valley or they keep climbing, right?
But at one point of their time, if you ask them this question, and we don't know when.
So a lot of you who are hearing this right now may say no to answer because that question, that answer to that question comes when there's a realization.
And that realization can happen at 80 years old or 80 years old.
So it may not come to many, many people as in, Saying, no, no, no, it is all worth it.
But at one point of time, when you ask the question, is it worth it?
What worth it is is friends, family, kids.
Yeah, relationships.
Relationships, that's what is worth it.
So speaking about this, right, so you mentioned about, you know, you have to challenge yourself, be uncomfortable about stuff.
And these days it's all about AI.
I mean, if you can switch topic from life perspective.
I think AI brings a lot of anxiety to people today, right?
And also brings a lot of maybe existential crisis.
You know, some, for example, developers, we think we are going to have no job in the future in a short, you know, time.
So what's your perspective with all this?
Especially you have dealt with, you know, like software engineering, building engineering teams for quite a long time.
So what does this AI era, you know, give you perspective?
Oh, there are three perspectives I have.
I was talking about, I was reading and talking and people, so first perspective is engineering practices are getting reinforced.
That is first perspective.
People are talking about a spectrum in development and TDD and I also wrote a post about, it's called expensive thing.
My point is that people are trying to make sure that whatever is being developed is developed correct, right, and all the stuff, right.
So that is first perspective.
So I am happy.
that people are reinforcing the soft engineering principles which we lived and breathed.
We both are from ThoughtWorks and we have lived and breathed the whole Agile TDD stuff and we know the value.
And we know that many, many people, when we go, we would go and tell them, please write tests.
They say, no, no, no, tests will slow us down.
But that is coming now.
Earlier, I think we did test it.
developers because they are humans.
Now we don't trust agents because they are not humans.
But both developers and agents did the same thing.
They both do the same mistakes.
But earlier we did not verify and we verified in a very manual way.
But now the amount of code AI is producing, you can't do that.
So that is first thing.
The engineering principles are getting in place.
Second thing, this is exactly like a cognitive industrial revolution.
par with the industrial revolution which happened 150, 200 years ago.
Like think about it when steam engines replaced horse carts.
Think about it when hand looms got replaced with the mechanical or industrial looms.
What happened then?
What did those people do?
One of the interesting things those people did, industrial revolution did not bring unemployment.
It actually bring more employment.
That means on the sidelines, there are different jobs now.
Instead of putting the horse cart, you are actually here to learn to become an engine driver.
And then you need somebody at the end of the train as well.
And then you need a conductor.
Then you need many more things.
And then you get more efficiency and all this stuff, right?
Like, think about it.
When people got aeroplanes, did ships and cruisers stop?
No, they did not.
But they reduced.
But then there is a new industry got created and people transformed and moved and operators became better operators, all this stuff.
So this is industrialization.
What I am trying to say is that each of agents will reward us as builders.
The problem is lot of developers think they are just developers.
I have always debated that software engineering is a balance of art and science.
That means software engineering is not an industrial process.
That means we all are creative.
If programmers are not creative, they can't program.
Unless they're very dumb programmers.
But if you're good, and those guys will get out of the job anyway.
They will get let go anyway.
And they have to do something about it.
And that is completely different stuff.
And what I'm saying is may backfire, but it's okay.
But we are creative people.
Now, if software engineers are creative people, that means this is age of builders.
If it is age of builders, it's a boon for us.
We can build things at different speeds, different things.
So what it means, right?
Now what it means is very simple.
Think about, I'm trying to figure out a very bad analogy.
I have a very bad analogy, but anyway, I'm not giving that bad analogy.
Think about, we have a dog and we have a ball and we throw the ball and the dog goes and fetches and we're trying to eat something, but dog is so fast that it comes back immediately.
So we are not able to try it.
And then we take it and throw it very far.
And we try to eat one bite and it comes back.
And then this time we're going to throw really, very far, right?
Now the dog and the ball fetch, this whole thing is agent.
Coding agent.
So what happens now, we type something, do this, it does that.
But we have not thought about it.
Now we are reacting again, it's free.
Now I have to type something else.
Now I'm typing something else.
And that fetch and pull is happening, right?
Now what has to happen?
There has to be a paradigm shift.
See, in the software industry, it is not a people problem.
It is a management problem.
Think about it.
Earlier, our bottleneck was the people who execute, software developers.
And then the bottleneck was not the people who ideate or give the problem, like product managers, BAs, and all that stuff.
Now the thing has flipped.
Now, we have to create enough thought process.
We have to flip it on things and start thinking like, how can we create more specs?
How can we create more better PRDs?
How can we define correct?
Like one of the things which TDD works and the reason we had Gojek with like a 900 plus microservice is thousands of deployers and we had down times, but we did not have that much crazy down times.
We could do root-court analysis and people could deploy 10 times a day.
They're like easily at 2,000 deployers or 3,000 deployers a day.
And we still could do that.
We did not wait for weekends.
Because we were defining, all the time all of us define what is correct.
See, tests define what is correct.
What is the judgment which says, what is acceptance criteria?
When we define acceptance criteria, we define both things, right?
What it means when it works perfectly fine and what it means when it does not work perfectly fine.
We write both, right?
Yeah.
If I give a wrong username and wrong password, it should fail.
We have to write that.
While we write also that given a right username and right password, it should log me in.
So that defining that correct means it also encompasses incorrect.
Now given that, if you now think through my view, now we have to spend more time, more chunky job, so that agent go on for a long time.
Yeah.
So instead of saying, what did you do yesterday, the standard questions change.
What did we deliver yesterday?
And what did we deliver yesterday means, what did we validate yesterday?
So now there is a whole paradigm shift.
Now you're a product manager, but instead of having one BA and four Delibers, you have four BA and one Delibers.
And maybe 0.5 QA, or maybe no QA.
because verification has to go up to the level.
Then what would QA do?
They also have their work, but their work is now changing to becoming more QA automation people.
So the whole paradigm shift is flipping.
The builders getting more hand, more control, and executioners, which are like developers, are getting less.
If you are going to remain developer, Eventually, you will have to go up and get your better skill at creating a better judgment, experiences, and then you will be able to get more useful skills.
What you pay for, what you are getting paid and what you pay for is always a skill.
You never pay for typing speed, do you?
So now what has happened is typing speed and some of the structured skill is gone.
But this is not only happening here.
You don't define correct over here.
Now you have defined correct in domains as well.
Financial domain.
You have to define correct.
If you don't define correct, then you will have wrong financial audits.
Lawyers, you have to define correct.
We know that whole hallucination and whole legal issues and LLMs generated fake cases and all stuff because they're just token generators at the end of day.
So the more structured jobs, will increase the circle of skills and you have to learn peripheral skills.
And then you can, and you should be a builder, all of us are a builder.
So it gives a fantastic chance for us to actually look forward for doing that.
Instead of trying to resist AI and trying to resist agents and like, oh, we'll still do, yeah, you can create artisanal code.
I was telling somebody, I was telling one of my ex-colleagues, saying, remember 10 years back, we actually wrote every bit of that line?
Actually, I was telling you.
Yeah, we wrote every bit of that line.
Can't believe it.
And now I'm actually happy, and people should be happy about it.
They should embrace this.
Yeah.
I still remember writing, get a setter, that's like the most boring.
Every time.
Control is press, control is press.
Every time.
It's so boring.
It's so, so, so boring.
So yeah, so I was saying, so when expertise is becoming almost free, whether it is design, and it's happening every day, whether design, security, audit, everywhere, then what becomes expensive?
And my answer is judgment becomes expensive.
So we need to fundamentally start focusing on judgment layer.
And judgment layer is going to be...
very important for everybody like how agents are doing because if agents can do good things fast they'll do bad things fast as well.
Yes, exactly.
We all see it in the cyber security space, right?
You have hackers and you have defenders.
Yeah.
A few days back, I read this document where a coding agent went and destroyed the entire AWS cloud environment and said, oh, sorry, I don't have backups.
I should have checked before doing this.
Like, yeah, but now, now actually you don't have anybody to just clean it as well.
But one thing that doesn't help is actually all these big AI companies are saying, you know, it's like fear mongering in a sense, right?
Maybe in the interest of increasing their valuation, whatever that is, they're saying like, okay, software developers, we don't need anymore.
Now, going to legal accountant and all that.
So there's this, this perspective that, yeah, maybe I need to really switch my skill.
No, no, no.
Look, 88% of AI companies are actually adopting AI.
Adoption has increased.
But only 6% have successful projects.
So why?
Because, look, it is, okay, I'll tell you this.
I'll give you this example.
I gave this example about Linux many, many, many years ago.
I'll give the same example about AI now.
So when you get into Unix Shell, right?
And it's powerful and what you get is a blank screen and a prompt.
And then you're trying to install one package and you don't know how to do it.
And then you struggle.
Like, AI is like that.
You are given a very beautiful, amazing car.
Take any luxury brand.
And you're given a key, you start with a car.
And to start it because you touched it.
Starting a car and running it on a big field is very different.
than starting a car and running on a controlled, carost road.
So a lot of people, what they're experiencing right now is they got this car and they got this big field and they're running like anything and it's all good.
But when they get out of that field and they exit the field and come to main road, now you have to put a lot of guardrails.
Who is going to put those guardrails?
Only people who know how to put guardrails will put the guardrails.
So our job...
is changing from, so earlier what used to happen, we do cognitive job, right?
All of us, like lawyers and financial, everybody does cognitive job.
We sit and do.
We type and the machine does something.
We don't do heavy lifting job.
So all that cognitive job was about to build guardrails, always correcting, always fixing.
And that was, as a software developer, that part in our life was very, very small part.
In the beginning, like morning, you read your story, you read your card, etc., you start typing.
Now the whole day you are just typing, typing, typing, typing, typing, and following that thing.
But now, our job is that typing, if typing is going to, if, think about this, if typing is done by somebody else, then what will you do?
Then you'll think more.
My point is, it is telling us to think more.
It is allowing us to, getting free from that typist job to more.
creative tinkerer job.
So I don't think so that is going to go away.
Yes, job will reduce because typing capacity is going away.
So if you're doing that, then you are going away.
Like for example, if you are just a front-end developer, yeah, like two years old, three years experience front-end developer, yeah, because that basic stuff, finesse has come.
But if you want to create a UX, yeah, cloud design can do.
many things.
But if you really want to create products and do something, experiences, we are the one who can create.
Machines will imitate, but they'll imitate so nicely, it almost feels like creation.
But if you go a little bit further on that age and treat yourself as a builder, then you will suddenly figure out, oh my god, I have so many opportunities.
Yeah.
And people are not thinking like that.
People are trying to use AI at a very minimal level.
They're not trying to offload their job and think more.
I think I want to encourage people to offload their more job to AI so that they can do something better.
But if you always have been that guy who goes and reboots the server, yes, your job is gone.
I'm sorry to say.
So I think what you brought up a very good point, right?
So obviously we need to level up ourselves, right?
In the value creation, right?
No more typing as like a skill or pressing that button, but more like the builder, the creative thinker, you know, putting the judgment.
I think this is also something that is very important, right?
Putting the guardrails, define correctness, incorrectness, and, you know, the judgment on what is good.
Maybe some people also call it taste, right?
Because only humans...
I hope, only humans that can give the face, you know, opinion, right?
Rather than the machine.
So earlier you mentioned it's very hard for anyone to predict now what's going to happen with this AI.
But one thing for sure, you know what people should not do.
So maybe tell us what we should not do in order to navigate this time better.
Oh, we should not resist.
First thing is that, accept.
Somebody asked me and we were debating on the framework and all this stuff.
And I tweeted about it two, three days back as saying, hey, are you here to ship the framework or are you here to ship the product?
That's very common, right?
And when I ask people, a lot of times I ask this, and one of my friends also asks this, what is your job?
If I ask you, you're in company, what is your job?
They'll say my job is this, this, that.
And I ask again, what is your job?
Think more.
And few people reach a place where say my job is to have value creation for the company.
Whatever role you play, the problem is people mistake their role for a job.
If our job is to create value for a company because we are giving our time to them and that's why they will pay.
If we stop value creating, they'll not pay us.
Very simple.
So there are multiple ways to create value.
So that's where I think.
First, accept it.
Second, figure out multiple ways of value creation.
We all are very good cognitive thinkers.
We are one of very few people in the world who are getting exposed to this and getting disrupted by this.
People who are selling hawker center things, they are not.
They are happy.
People who are delivering food, they are happy.
They will get at some point of time in bots and drones will replace the food delivery.
or some other delivery, but still there will be more things.
So my point is this is a cognitive industrial revolution where this happens.
So first, accept change.
Second, think about your value creation.
And third, what not to do is very important, is get worried about it.
There's no reason to get worried about it.
Because unless you stop worrying about it, you can't think about something more.
Maybe I'll not get a job.
It's okay.
Can I bake a bread?
Maybe yes.
But I have this responsibility, I have that responsibility, I have that responsibility.
Fair.
If you don't do anything, those things are still constant in your life.
As you said, if you think about yourself, those things don't change.
If worrying about things can change things, then I would worry whole time.
And I'm culprit of that.
I worry about my son, my daughter, not daughter, my son, my wife, my relatives, my friends.
I worry about them.
But that worry gets away.
Worried about them or thinking about them like somebody falls sick, we worry about them.
But we pray.
Not that religious pray, but a spiritual pray.
So my point is there is no reason to worry about it.
If you stop worrying, then you can think about path forward.
A fourth thing which I think is very important is that stop running.
Close your eyes.
Stop looking at your device.
Stop getting consumed by stupid internet.
Give yourself time and then think how beautiful the world is around you.
And then you will think about what can you do.
That concept many of people will not agree with me.
They will give me excuses, reasons.
Real reasons saying, hey, you're talking about that, but I have to pay bills, I have to do this, I have to do that, and I just got laid off.
There is no way you can get back to the same company.
And it's unfortunate.
But now you've got a time.
You've got a forced time.
And if you stop worrying and stop, a lot of people just start immediately applying for a job somewhere else.
That used to happen earlier.
But the problem is you got laid off because there is some skill which got automated.
If it's got automated company A, it will get automated company B as well, and company C as well, and company D as well.
So you can apply five places, and if you are not a value creator for them, they're not going to hire a typist again.
This is a very crude way of saying it.
I'm so sorry to say this.
So my point is people should actually learn.
Actually, you know what, Henry, I'll tell you this.
Don't you think our job hasn't always been, our industry has always been out learning.
Like when you started, maybe you are working in Java or C++.
Now you work at Python, Rust, Google.
You learn on the way.
I think people forgot that our learning exposure or learning attack area or exposed surface area has to be much larger now.
So, for example, I'll give you one more example.
15, 20 years back you could go to a GP and not go to specialist.
But today, if you've got a tooth problem, you'll go to dentist, but you'll go to the best dentist who has done masters and postgraduate, everything.
If you were to go to kidney or whatever, you'll go to MD and whatnot and a bunch of things.
I think that also, I was telling somebody, that also has come to software engineering.
Now you just can't have a comp science degree.
You had a comp science degree, then you had to do MS, then you had to do something more, because this value creation has just gone really high.
Yes, it has gone very high in a very short amount of time, and people are reacting to it.
But I think eventually it will be fine.
There are different kinds of jobs that will get created.
For example, people actually are not good at prompt engineering at all.
That is a new job got created.
And people are not good at that at all.
People should be good at that.
Actually, there are papers about prompt engineering.
I was reading a paper from Google.
We are saying, if you repeat the prompt, LLM does it better.
Something like that.
Yeah.
Now, that is about learning, right?
Yeah.
So people are not doing prompt engineering.
A lot of people just use cloud code and just use cloud co-work and codex and all stuff.
But they are not thinking through, how can I automate the whole stuff?
I don't know how many organizations actually have built and put agents on their GitHub repositories to auto-merge the PRs.
First of all, PRs shouldn't be there.
But that is different.
That is different software engineering.
Talk about tongue-based development and all that stuff.
But have they put PRs?
Like, how many of them?
Whatever people talk about automation, you know, they're talking about automation on their laptops.
Yeah.
They're not talking about automation at the...
central level.
They're not talking about saying, hey, we will not develop our laptops anymore, we'll just do a cloud-based workspaces and work over there.
How many people are doing?
Because once I already works on the same box, same machine, shared terminal, then you can have agent which can look at everybody's source code.
How many people have done that?
That is a skill they should be doing.
Like remember, we went through this DevOps thing, right?
The thought process was that we don't need Anybody else and developers should be doing operating.
What happened?
Other companies which could not upskill the job, they created DevOps role, man.
Yes.
I hate that.
I don't like it.
I don't like it.
Because DevOps is set of practices.
It's not a role.
But now, big companies are advertising it as a role, and there's a DevOps role.
I'm like, no.
That just sounds so wrong.
Yeah.
So that's where it is, right?
So my point is, you have to now, your full stack engineer definition has changed.
Earlier, full stack was front-end, back-end, and whatever, whatever.
Now it is like front-end, back-end, how to use tools, and how to write agents, and you have to do that.
That's it.
Yeah, so I think really good point.
Don't resist.
Do not get worried.
Think about value creation rather than...
you know, your tasks, your day-to-day job, right?
And I think participate, right?
You need to learn about all this AI stuff, you know?
Otherwise, I think you'll get replaced or you'll get irrelevant in the world, in the industry, right?
So speaking about engineering practices, you mentioned it is getting reinforced.
And I think for people who wants to leverage AI in the right way to always produce correct things, you really need these practices.
Maybe if you can...
give you a few top engineering practices that needs to be practiced really so that we can leverage and benefit a lot from this AI revolution?
Oh, so there is a book called Philosophy of Software Design by John Osterhout.
Yes, I love that book.
And there's a paper, 1975 paper called No Silver Bullet.
I love that paper.
Yes, I love these two things.
And now then there are books around patterns and software engineering and enterprise architecture, all that stuff.
But it kind of comes down to like three or four basic principles.
For example, software is about composition.
That means you have to be able to build software layer by layer, brick by brick.
That means abstraction.
So first one is abstraction.
I always say abstract everything out.
Like when people ask me, When sometimes I ask people, what are you building?
They'll give me a very big fancy thing.
We are building this, this, this.
And I say, no, no, no.
What are you really building?
And if you go down like two or three or five hives, basically five hives, if you go through that, finally they'll figure out they are actually building some CRM or some CMS or some SKU listing thing or some payment checkout stuff.
You put them all together and this becomes a fancy product, which is great.
Software engineers should think like that and we should now who are builders, they should think like that in terms of abstractions and build thing.
Like build one thing which does only one thing and then build one thing on top of it like Linux commands.
If you want to count the files on a directory, you say ls minus l, pipe2wc minus l in your DOM.
So one is that.
Second is about define correct or like domain-driven development, behavior-driven development, test-driven development, whatever you want to call it.
do that.
That will give you confidence in the software produced.
Third, now I never liked PRs and I don't like PRs still today.
There's no reason to do PR because now more so because my agent will generate like 3,000 lines in a day, your agent will generate like 4,000 lines in a day and we both do a PR and then we'll pull our head and their agents will pull our head and do absolutely wrong things.
to merge those PRs.
It just becomes a race.
Who checks in first?
Henry checked in 530, I checked in 530, oh my god, I should have known when he's checking in right.
And this PR merge help companies are still going through that, but they don't realize it.
So now, trunk-based development, do a branch, agreed, put autonomous code, merge the branch, but don't do PRs.
Doesn't matter, nobody's there to review.
You are the one who is there to review.
So trunk-based development, go move towards that.
you do get commit a hash when you do a check-in.
So you can actually pick your hash out and do something, right?
Fourth thing, feature flags become very important.
Super important now compared to any time.
Check-in your feature with a feature flag, check-in disabled or enabled based on your environment.
So feature flags.
So we talked about abstraction, we talked about spec, trunk-based development, and then feature flags.
Fifth thing, which is very important, is that try to run the whole damn test suite.
Like bring CI.
A lot of people use CI and CD as a buzzword.
I want to again reiterate, CI and CD are two different things.
But because of industry and a lot of things, we say CI, CD.
Continuous integration is very different than continuous delivery.
One, validates your software and gives you a produceable build.
Many people have forgotten the concept of golden build.
Bring that concept back.
And then CD, continuous delivery, is about taking that golden build and deploying it on your server.
CD does a last mile delivery and CIE does a first mile verification.
So these are the four or five things.
If people do that, I think that is good enough.
And then there are many more things.
But I live by these and these are the easier ones to actually crack.
Yeah, yeah.
So I think thanks for sharing all this for listeners who are curious how to actually excel in this AI era.
Some of these practices actually has been around for years, decades.
But somehow we always tend to either forget, miss practice, or whatever that is.
So I think thanks for sharing that.
I know that you're also at the site building this, the claw thing by yourself.
Yes.
Some people call it vibe coding, some people call it solo developer, whatever that is.
How can you succeed actually building something?
I assume it's also not trivial, right?
It's not like to-do list or something like that.
It's actually trivial.
It involves infrastructure and some kind of reliability thing if it doesn't work right.
So how would you advise vibe coders out there?
Do they actually need to practice all this as well by themselves?
At least first three.
The first three are the very important, which is I think abstraction, specs, and the trunk-based thing.
I think that is very important, and I think that's how I built it.
You have used it, and you're using it.
And that is a little bit of a complex piece of software because we dynamically procure infrastructure, then we set up a machine, then we...
put your open claw there, we have baked image, then we take your parameters, we set with the open claw, there's a proxy, there's a dynamic DNS, so a lot of pieces, moving pieces.
But the way I built it was very simple.
I wrote about it, it's called the Small Island of Delight, what open claw is.
And then I wrote next thing, what 10 walls all developers feel and face.
So basically the way I built it started very small.
First one was to just automate.
open claw installation on my machine so I can give that a script to my friend and they can do it.
But then once you've done that, then you have to make that machine, make the script execute using a web trigger.
So you write a small web application.
Then behind the scene you write a small worker.
But now that worker needs to know which machine.
So you put a job in the queue and it goes to make API called the cloud and get a VM for you.
And that's how I built it.
But what I did, I spent a lot of time, like the story cards, I'll always carry a bunch of story cards.
And I wrote, literally I sat down and wrote the stories.
And then I, and we know that we do this stacking of the story and find them out, like what are dependents, what are independent.
I did all that stuff.
To start with, I had like first, first day I had around like five cards.
But next day they became like 12 or 15 cards and then now we have like 300 cards.
We went through that.
But then what I did, and I'll tell you about that as well, I did a role play with the agents, but I used issues, GitHub issues as the anchor.
So I'll get a PM agent to read my description of card.
and then create a detailed card.
Then I'll get a tech lead agent to verify that card and I would use different clock provider.
And then I come back saying you got as a, and then that I have shown you that the PM agent actually, tech lead agent actually puts a comment on it.
Then I get one more agent saying, hey, you are the developer, how would you implement, put the how as well.
And then we learned the whole thing.
And I have defined my agents at MD, which talks about SDLC cycle, what they should do, what they should not do.
Always, always deploy on staging.
Always.
Never deploy on production.
I deploy production.
I never ask Claude to deploy.
I always click that button myself.
So I had to click that button 15 times a day.
So I set up a CIO over there.
So that's how.
Very small.
But I did not write code.
I did not write code.
That is amazing, actually.
I'm very impressed with that.
So what makes it possible for someone who can let go writing code?
Because I see these extremes, right?
Some people, even in big companies like Anthropik, Spotify, they say no developers writing code anymore by hand.
But some people still think it's not possible.
AI hallucinate.
There will be incorrect code.
What to do about it?
Maybe you can share your experience.
How do you actually...
kind of like let go this, you know, having to deal with code?
First, I actually, that's when I, honestly, that's when I realized that judgment layer is very important.
First 10 days in February, we are nuts.
I was trying to make it work and because like the way, and of course I improved.
I was told you, first day five cards, 10 cards were there.
But those 10 cards didn't have anything to say, how will I verify?
What process I'm telling you right now, after like so many days and so many iterations, going through like 300 cards, right?
And then one day it stuck to me.
I'm not going to do this way.
I'm going to do my way.
I'm outsourcing typing.
That's how I did it.
And I tested the system and all this stuff.
So that check and balance.
So people who are not ready to let it go, think about it.
if you're not typing, then what would you be doing?
And that's what I thought.
And if I'm not typing, I'm, well, that guy or that it thing or it or a guy or girl, I don't know what it should call me that.
But it sometimes is very amazing, huh?
Like it talks to you.
Oh, I'm sorry, I did not do this.
It's like so amazing.
Sometimes like, oh, wait a minute.
We should have done that.
Like, okay, fine, whatever.
So put it in sandbox.
First thing is that, put it in the sandbox.
Have your...
exact staging production-like environment.
So I have two staging environments.
I tear them down and bring them up.
Do automation.
Actually, one more point is very important.
It's do automation.
Automation is very important.
Do automation on all the small things that you don't have to do.
So in my software development lifecycle, only automation I don't have is where I go and deploy.
Apart from that, everything is automated, right?
So given that, if you do a sandbox, if you...
abstract your small things out, if you define correctness, if you spend a lot of time on judgment layer, I think it does not hallucinate.
Now, at least.
Yeah.
Yeah, I think the model has become so advanced now, and especially if you have a good harness tools, right, like code code and all that.
I believe, yes, hallucinations still happen, but if you build this correctness layer, right, your spec, your TDD, the things that can verify the behaviors accurately, I think, yeah, this thing tends to produce Accurate.
And I am very thankful to all my users of ClawStations.
I don't have hundreds and thousands, but whatever I have.
Whenever they ask me something, I'm able to actually respond to them and fix their stuff, whatever it is.
And it's pretty interesting because I have exact replica of production running with me.
So that's where I am.
Heads off to you.
I don't know how you do it.
One single person building.
I think it's quite complex to build something like a claw station.
And yeah, I've been playing around with it.
So thanks for creating that.
One thing also you mentioned, I mean, not just software developers.
Organizations are going through this moment as well.
And they need to think how to implement, apply AI in their organization.
One thing very interesting you mentioned earlier, right?
So many people just think of automating, you know, maybe something on the local or maybe...
tasks that they need to do, right?
But never thinking holistically in terms of maybe putting agents centrally and all that.
And statistically, I've read also in some articles, maybe 95% of organizations that apply AI doesn't actually reap the benefits.
Like there's no ROI coming out of this.
Not yet.
So tell us what's wrong.
Like how come?
I'll tell you what's wrong.
Around 2000, year 2000, a lot of a lot of organizations started moving towards electronic forms.
And I'll give you a simple example.
So electronic forms getting, I'm your customer, right?
Earlier you had a physical form to get complaint.
So what will a physical form will have?
My name, I have to fill my name, my date of birth, where I think, and I put a complaint over there.
Now what happened in early days, those forms became digitized.
But I'm your customer.
So who designed that form forgot that they don't have to ask me where I live every time.
Because when I log in and set up my profile, that information is already there with you.
So if you look at earlier forms, a lot of information were again repeated again and again and again.
And I'm like, why?
And this happens in a lot of forms still when you get bank forms and all stuff.
Those bank guys know my name.
But why I had to fill it every time?
Why I had to do DocuSign every time?
Why can't they do something?
So that digitization took time.
Like we are living in Singapore and I actually see Gowtech has done some fantastic job around that, right?
Yeah.
Like whole SingPass, whole ICA and all this, it's amazing, right?
So given that, but that thought process came later.
Earlier what was?
Oh, it is a physical form, let's capture this electronically as it is without rethinking the workflow.
That's exactly what is happening right now.
Now we have smart, reasonable reasoning and thinking agents which can produce, but let's not change anything else.
So we have a trick or technique called value stream mapping which you know.
A lot of organizations are not doing that.
They're not doing value stream mapping.
They're not thinking, rethinking, saying, if we keep AI in the center, then how would we behave?
And that's where we start thinking about, how do you behave smartly?
What can you offload?
What you shouldn't offload?
So people are not doing that.
Till the time people will keep using AI as a, what is it?
I'll give you a simple example, one more example.
You have microwave at home.
Yeah.
How many buttons do you use on that?
Maybe just one, two, two, maybe max.
How many buttons it has?
A lot.
A lot more.
I mean, like maybe 10 or 12.
And now they're on internet.
Like, oh, yeah, like you can start micro using your phone.
Yeah, but I have to go and still put the thing inside the micro, right?
There's a joke by one Indian comedian, and I love that joke, right?
And it's same like this.
Like, if you can build, you can have all the features, but if you don't keep people in mind.
then you can keep building useless features.
So sometimes it is about reducing what you already doing.
For example, the physical form always had the profile or user and name, address and all that stuff.
You have to reduce that because you don't need that.
Same with AI.
You have to reduce a bunch of things.
You don't need those.
So now you have person one, he does something, then person two does something, he, she does something, and then person three does something.
She does something.
Yeah, actually have person one, two, three sitting over there and doing the same thing, but they are doing it with use of AI.
But they are still passing the buck from here to here.
While actually person one can do this thing with three agents.
Why we are not doing that?
So that is where they are not understanding how to keep the AI in center and organize organizations, workflows around it.
So one of the most important things people have to think through.
Any organization has, only thing they have is workflow, workflow, workflow, workflow, when they're serving somebody.
They just need to think of that.
And if they can do that, then things will become better.
For example, our most popular software engine workflow is this.
Ajay works with Henry.
Henry's Ajay's tech lead.
Ajay's junior developer.
Ajay...
creates a code, Henry gives a PR, Henry reviews the PR as if Henry has a compiler fit in his eyes, never checks out, says, OK, it is good, or puts a comment.
And then Henry merges the code, and then it goes to CI.
Now, if I am an agent, why do you need to do that again?
What you can do is you can tell me, as an agent saying, Ajay, do this, but once you do this, and once you run this, make sure this, this, this happens.
or tell somebody else.
Two agents, they can run.
You can have five agents on your map.
But we have not done that.
We're still creating branches with agents.
Can you believe that?
We're still creating PRs with agents.
Can you believe that?
Why?
If I am the verifier and if I am the executor, then why am I doing this?
So they have to, like, they're still using it as a minion rather than the builder.
Like, execution is gone.
So let it go completely, you know?
So that's what is happening all over the place.
Yeah.
So I think definitely this requires, I don't know, paradigm shift.
Thinking differently.
Coming back to value stream mapping and all that.
I think the intention is not to automate what you are doing now, but also holistically thinking, how can you let agents run by itself, natively, as if AI is the center.
For leaders out there, I think many are thinking, AI means I can reduce a lot of people layoff.
Any thoughts about that?
Because I'm sure, yeah, maybe in the beginning you can do a lot of efficiency and all that.
But what is the risk?
What are the dangers of doing it this way?
I think if you apply our way of doing things in software engineering, you can already reduce a lot of people.
Yeah.
Okay.
It's not about people reduction.
It's about...
basic job going away.
So, organization can certainly reduce people, but they will have to certainly hire more people as well.
Like, how much ever people are doing hoopla about this, there are still people hiring.
Why are they hiring if the people don't need software developers?
Like, big companies are still hiring for different reasons.
Right?
So, before reducing the people, It is not like that the AI is going to do five people's job.
It's not.
Again, that's my thing.
Management has to think through this.
It is flipping on its head.
Yes, you don't need five developers.
You need only two, maybe.
But now you need, instead of having two BAs or two product managers or two project managers, you need four or five.
Because now you can execute five things in parallel.
So what happens?
How are you going to do that?
What will happen is you will actually reduce your output that this person...
is still not able to produce more things for this person.
So, organizations, yes, you can certainly reduce people on the very basic structure.
See, the thing is very simple.
What is getting replaced?
The most structured things are getting replaced.
Programming language are the most structured.
So that's why LLMs can reproduce them very much better.
That is the first most structured.
Second most structured, maybe finance.
Third most structure, maybe insurance.
Then fourth, little bit subjectivity comes the law.
But still there is some...
So anywhere the structure is...
Structured work is getting replaced.
Cognitive work is not.
The organization should...
If they are reducing it, it's fine.
But they are creating cognitive gap.
They can't live in that vacuum.
So they need to think through that.
They are not going to go faster.
They are going to go same speed.
That's all.
If they are okay with that speed...
Yeah, let people go.
It's fine.
But if you want to accelerate something, then instead of letting people go, think through and enable them with more skills.
Yeah.
Yeah.
So I think that's a very good point, right?
Instead of efficiency only, right?
Think about how many innovations can you do if you upskill these people, let them build more, you know, innovate more.
And I think, yeah, especially now with AI, many other companies, small teams can actually disrupt.
you know, the income ones.
Yeah, and tokens are expensive, man.
People don't understand tokens are expensive.
People are burning a lot of tokens.
They're expensive.
It's not a pride.
It's not a pride.
Yeah.
Sometimes you don't have to have tokens burning machines.
Yeah.
And it's okay.
So organizations have to think a balanced approach.
Yeah, and especially now we know that...
tokens are still subsidized because of a lot of investments.
Yes.
And people are still using, people want to use Opus 4.6 max effort for writing a getter and setter.
Why?
So yeah.
Yeah.
So what do you think the implications of AI also for Southeast Asian?
Because we know the models all are coming mostly from Western.
There are some open models from China.
But for Southeast Asia, what do you think the tech scene is going to?
We have always been consumer, by the way.
We have not been, in terms of software, we have always been consumer.
Not only AI models, like my watch, your watch, my phone, your phone.
Everything around us is either Western or Chinese, right?
Because those are the hubs and we have always been consumer.
We are doing some good stuff over here.
And that will change.
But...
If you look at Southeast Asia, the economy is mostly non-tech economy.
It's mostly financial and other things.
So there will be some automation, but not a lot will happen.
And the non-tech economy, the economy with more human dexterity will remain a little bit unchanged for now.
If you have jobs which has human dexterity, your job is okay.
If you're a physiotherapist, you're fine.
The other robots come in.
No, no.
I'm not going to put my head in like a robot barber.
Like there are like scissors over there, man.
What are you talking about?
I'm not going to put my head over there.
No, no, no, no, no, no.
Once it goes wrong, I cut your ear.
I'm sorry.
I'll do it better.
I'll cut next time.
I'll cut both.
No, I think our industries will get more modern.
I think it's good for us.
I see it as a boon for us.
It is a very good boon for a lot of entrepreneurs and builders.
My view is that people who have been thinking about building, this is the best time to build software.
It is the best time to build automation.
It's the best time to put, like you want to do a dropshipping and you did not know how to build a website and payment system.
It's the best time.
Just go ahead and do it.
Like put 15 dropshipping websites on the thing.
Something will sell.
It's a good time, man.
I see it.
And since we are a consumer economy, we will get benefit out of it because the AI adoption and all this stuff will make our experiences better.
There will be many people who will actually work on building those experiences for us.
And there will be better products.
So I think it is a good chance for us.
It's a good chance for all the people who have been...
thinking that West was, or China, or Japan, or Australia, or UK was the advantageous place and no level playing field, it's a level playing field for us now.
I'll tell you 15, 20 years back when I used to go to US, people used to ask me, can you bring this, can you bring that, can you bring that?
Now nobody asks me that, because everything's at level.
Same thing is gonna happen with the technology.
Now we have very good chance and advantage to create things over here.
and give it to our consumers and go to global scale as well.
Like if you want to use software, you can use my software anywhere in the world.
Yeah, why not?
Right.
So one thing that like some other people also think about using AI a lot more is that we're actually not learning a lot because almost everything now you can ask and you get the answer.
Like fundamentally, you know, in terms of learning, in terms of going through that hardship to actually, you know, understand.
principally what is going on, I think it's getting reduced because it's so easy to just do something.
We're talking about creating stuff, let the agent do that.
What do you think people should do in order not to lose this aspect?
They should not do that, basically.
I have found very few people who actually learned piano using YouTube.
You still need an instructor.
You still need a coach.
So I have found if you are trying to do things, the more far-fetched away you are, at some point of time, you will hit your limit, and that point of time you will be redundant, especially in cognitive industries, especially in knowledge workers, not anybody else.
So I think people should not, that's what is happening actually, if you look what's happening with software developers, because they stopped learning somewhere, they're like, Oh, let's do this.
But they did not learn how to understand and learn about models and how to use the models, right?
So that is what it is.
Earlier we thought, no, that is all is going to stay somewhere else and we'll do AI, we'll do this, models, we'll do this.
They're doing ETA predictions on Google Maps and all this stuff.
But now, same thing just came in and came in very fast.
It's a very fast change.
Next era is going to be about back to services industry.
SAS will get disrupted, but a small SAS will get disrupted much faster.
A large SAS will also get disrupted.
So it is excellent for services industry.
If you want to remain relevant, you have to learn.
And I think basics, if you learn basics, then those basics can be applied anywhere.
If you learn programming as a basic, so I tell people sometimes, you need to learn only four or five things in any programming language.
How to define?
A variable, how to call a function, how to run a if condition, how to run a loop.
If you know this, you can learn any programming language very quickly.
But a lot of people start with some other language, but they never learn the basics.
They know how to do that.
But then they switch to a different language, and that language sounds very strange to them.
But if you learn the basics, then you can switch languages.
Same way if you learn basics and don't forget about reading and being curious and being passionate.
So I always say there are three things about us as an individual which makes us very valuable.
We have to be responsible.
That's our job to deliver something.
We have to be passionate.
That's where we contribute to not only to our job but our teams and everything else.
And third, we have to be curious because that enables us to go to first two.
So curious, passionate, and responsible.
If you have to continuously go through this circle, then you're fine.
People actually, I think people who have been curious, passionate and responsible, they aren't affected.
I think so.
Wow, I love that.
Being responsible, being curious, being passionate.
So definitely key learnings for all of you who listen.
We have talked a lot.
Is there anything else that you think you know you have?
No.
Thank you for hosting me.
It was a good conversation.
I had no idea what we were going to talk about.
But yeah, looking back.
My perspective has changed a lot and I wish a lot of people get better perspective in life and see what goes on.
It's still learning, man.
It's still learning.
Right.
Yeah.
So I don't know whether you still remember.
Every episode I will ask these three technical leadership wisdom.
So I know you have thrown a bunch of wisdom.
So maybe if you can just give three version, you know, short version, what wisdom that you would want to leave the listeners today from this episode.
I think life perspective changes with the age.
Right.
So when I was 25, I would say something different.
When I was 35, I said something different.
And when I was 50, I'm saying something different.
So first thing is that when somebody offers you some help, never say no.
Because their intent is always to help you and make you better.
So that is what I have learned.
That is one wisdom.
I'm talking very different.
I would have talked something very different.
Now I'm talking very different things.
So first one is that.
I think second, which is very important, is like in life, the more you age, the more you should and you will value friends and family because you have very few weeks with them.
Like, for example, if I'm going to live till like 70 or 80, I have only 1,500 weeks.
That's not much.
Weekends.
If you talk weekends, oh my God, only 1,500 weekends.
And third thing.
What are we talking about?
Be curious, be responsible, be passionate.
I think that is very important because that will keep letting you go through the life.
And then there are a lot of regular stuff, but yeah, I think those three things are one.
If somebody offers you help, don't say no, just take help or ask for help.
It's okay.
Value your friends' families more than anything else.
And third, be curious, be passionate, and be responsible.
Lovely.
Thank you again for this conversation.
Ajay, I'm always excited whenever I talk to you because there's so many wisdom that you always gave right through your stories, experience, and also examples that you also live by.
So again, thank you so much.
And I hope people learn a lot from this episode.
Yes.
Thank you so much for hosting me again.
It was amazing, amazing talking.
It was great.
Thank you so much.