Who here thinks that the job won't be automated by AI in the next five years? Wow. What do you do? How do I get a job? So, who here thinks that your job will be automated in the next five years? Okay. So, I just did this question yesterday. It was just curious like what people online would think and it seems like a lot of people think that the jobs are going to be automated.

Recently I launched something as a side project. It's small. It got like 300,000 views. And within a day, I got an email from someone saying like, "Hey, I love what you built. So I use Clico to recreate exactly that and here's a link." And I'm just like, I'm flattered, but also like what? So it made me realize that whatever exists can be replicated. I just fluctuate between excitement and despair because on one hand, I feel like now I can build anything I want but at the same time anyone can build anything I want. So what is the incentive structure for me to do anything?

I have stopped using a lot of what I call like SAS light. I feel like it solves a very small problem and charge me like $50 per seat per month. Why should I keep paying for you when I could just recreate exactly that with AI? People tell me data is a mode but it turns out that data is not a mode, it's just like expensive. It's not really a mode if somebody can just throw money at it and acquire data and build it. I call it like the Ghibli moment of software. If you can describe a style, AI can generate it. If you can describe a software, then AI can build it for you. You remove the need for imagination.

I built because I want to solve a problem. One thing I do believe is that there would always be problems to solve. Problems follow a long-tail distribution. AI would get really good at common issues, and over time it would cover more and more edge cases, but the edge cases would never go away.

Human preference is one thing I don't think that human preference is just an equations that people can just package nicely. It's very personal, very culturally dependent, geographically dependent, age dependent. I'm from Vietnam. Here, companies deploy customer support chatbot usually go the text route first. But in Vietnam, people are on the motorbike all the time, so they actually really don't like typing. So companies in Vietnam actually deploy voice bots before they do chatbot.

There are a lot of cultural nuances. For example, the response time. In the US, people expect you to respond instantly, like 80 milliseconds. Whereas in Asian culture, we actually wait a lot longer, more like 200 or 300 milliseconds to make sure the person finish. Voice bot you want to balance latency and also humanness. You have to understand all these nuances. Only when we go into specific use cases and specific demographics we're targeting that we can understand.

Humans interact with AI. People are trying to retrofit whatever exist to fit what they think is a new workflow. What's a fundamentally difference between a terminal and an IDE? A lot of that is still ongoing questions. Human to human collaboration is actually very complicated. Who here is still review every single PR line by line?

I talked to a team recently. The senior person reviews team members' AI generated code line by line for education. But junior members were like, I'm not writing the code, how do I give that feedback to my AI? The whole workflow of reviewing code is very outdated. Senior members should be giving feedback on how you give instruction to AI to produce better.

AI interaction with the environment is another thing. Terminal is already getting a bit more dangerous. Recently Claude just wipe out my Postgres locally because it was trying to create a new app and that port was taken. A lot of environments that AI currently operate in are reversible. You can revert back to the last commit or previous backup. But there are a lot of environments where the action cannot be reversible. If an agent fill a form and click submit, we cannot reverse that. Or in the real world, a car runs over a pedestrian, you cannot reverse that. We need to build out the guard rails for the non-reversible actions.

Who is making the world more agent ready? Websites, rate limits... the whole concept of rate limit is quite irrelevant to AI. I did a benchmark on how AI do web search. It visited a thousand URLs but only 20 were unique. It's a very human way of doing search. There must be a more efficient way that are less human-centric and more AI-centric.

The question nowadays is usually less about how to build. The question is what you build. Who is going to build the things that don't exist yet? Who should be imagining it? Fundamentally I enjoy building, it just bring me joy. I hope we can normalize building things for fun. Before we had mass manufacturer clothes, now people want something custom made. I'm not sure if we'll have a future where there's "artisan software", but for now, I enjoy building apps as a gift for my friends. AI does make my life for now happier.