# Transcript

**Courtney**: [00:59] Hi. Good morning.

**Anand**: [01:02] Good morning. Let me just give it a minute and start off.

[Pause 01:04 - 01:40]

**Anand**: [01:41] Okay. Why don't we dive in? But before we go, any questions that any of you had on what you were working on yesterday?

**KK**: [02:02] Sorry, can you repeat?

**Anand**: [02:03] Any questions from yesterday?

**KK**: [02:08] I don't really have any.

**Anand**: [02:15] And if something comes up, let's go ahead. But let me then dive into what I wanted to check. Were you able to show your work to a couple of people and what did you see? Anyone?

**KK**: [02:35] I guess I'll start first since my mic is already unmuted. For me, I didn't really had—okay, I know the email said two students, right? But I didn't really have the chance, so I went for my parents. I don't know if that's okay, but I mean, they played around with it a bit. I think the most impressive thing, **the "wow" factor, was like: "You did this within one hour?"** But otherwise, it was like: "Uh, okay."

**Anand**: [03:11] Which is interesting. And others for your context, this was a news summary application. So KK, could you please the link to that on the chat so everyone has a copy of that and they can take a look at this as well? And yeah, that was what I thought as well, which is: "**Wow, okay, this is great. I can get a summary of the news very quickly. Then... so what?** Or what do I do with it? Would I really use it?"

[Echo interference 03:45 - 03:57]

**Anand**: [03:58] Perfect. Thank you. So, that gives us a direction. And I think that's exactly the kind of useful feedback we need, which is: **once we get over the "oh, this is really impressive," what do we do next?** That becomes part of the question. What are your thoughts?

**KK**: [04:22] Like, such as what exactly?

**Anand**: [04:24] I mean, you built something quickly. That's great. You showed it to your parents and they said, "Yeah, impressive, but..." So how do you make this a product?

**KK**: [04:41] Introduce monetization?

**Anand**: [04:45] Interesting. So what you're saying is: I've got something, I'm not sure if it's useful, so let me charge for it.

**KK**: [04:53] [Laughs].

**Anand**: [04:54] Give it a thought. I think it's worth exploring. But **the good part is finding problems is hard. Being able to verbalize problems is hard. Once you know what the problem is, AI can help you solve them.** Part of it is just about redirecting it. Let's see. Okay, let's take another app. Any volunteers? Anyone showed it to any others?

**Johan**: [05:43] I can go next. I made the app as a product for my parents because I think it would be useful for them. I received some of their input and I think that it may be still a bit too confusing at times where there are features that are not very needed, while the essential ones that they find useful are not included inside. So, I think I know how to iterate next on what.

**Anand**: [06:21] Awesome. Let me share a link to your app on the chat window. And this was, as I understand it, a walkthrough of stock market tickers, right? Are these commodities or what are these?

**Johan**: [06:37] Would you like me to share my screen?

**Anand**: [06:38] Yes, please. That'll be great.

**Johan**: [06:53] So, I'll try to explain it as best as I can because it's in Indonesian.

**Anand**: [06:58] Sure.

**Johan**: [06:59] Oh, maybe I can use the translate feature. That works. So, I wanted to pull real-time data from Yahoo Finance and it seems that it successfully did that. And there's also a feature to download the past history of one stock—each of the stocks by the emittens (issuers) from the website and you get a CSV. And here is a tracking tab, where it tracks your purchases, how many stocks you own, and it can calculate automatically the profits or the margins that you will get after you sell it. And this is the average price and the break-even point of the stock.

**Johan**: [08:21] "Transaction Auto-Reject Limit." This is not very special because in the stock applications these days you can set an automatic selling point. So if the stock reaches that point, it sells it for you through the application. While this is a tracker, so at best it will just remind you to sell it; it's not as instantaneous as the applications these days. And this is the journal after the user submitted how many lots they bought.

**Johan**: [08:55] And "Saving Plan." So, this is the one thing that is quite novel, I believe. **Instead of having to consult a financial manager in the bank, you can ask it to help calculate how much you want to save, so you don't go buying and selling very quickly in a short time span.** And maybe you can achieve your financial goals. And this is a dictionary—which is funny because I told it that I was a beginner in stock trading and then it added this in, which I didn't give it any specific instructions. So this was a surprise to me. That's all.

**Anand**: [09:48] Interesting. And you had multiple turns of conversations with Claude, is it?

**Johan**: [09:56] I had one extra... I told it one extra thing because it said it omitted A and B and I told it to add it back.

**Anand**: [10:11] Got you. And you said this is something you thought your parents might use. What was their feedback?

**Johan**: [10:20] The essential points... they also use AI to help with their trading, but the essential points like reading charts (which improves efficiency) and finding predictions—which AI rebutted on because it says it's not a licensed financial advisor so it cannot give predictions. But if you use the generic model of Claude, for example, or ChatGPT, it will still give some form of predictions based on calculations. And I think that this lacks those mathematical calculations that are required when one is purchasing or selling stock.

**Anand**: [11:13] Got you. Would they use it? Would they use it if you made some of these changes?

**Johan**: [11:18] I think so. My mother especially is a fan of Claude.

**Anand**: [11:24] Fair enough. Give it a shot then. Because there are two parts that I would encourage everyone to think about. **First is, sometimes when we are getting feedback, it's easy to miss what's being said.** So, I usually do these on Teams sessions or Google Meet or whatever so that I can capture a recording of it, possibly with video, and pass it to Claude or ChatGPT or whatever. Say: "Here's the feedback that I got." If nothing else, just write down everything. I may have remembered ten things; it may tell me five that I forgot.

**Anand**: [12:13] **Second, I can short-circuit my thinking process and say: "Here's the feedback. You fix it."** In other words, the users start shaping the application directly instead of having to go through you. And if there's some value that you can add on top of it, you can always do that, but that transcript has a lot of value. One of the things that I tell the interns who join Straive is: **"Get into meetings, record the meeting, don't even try and think about it because you won't understand half of what's being said. Just take the recording and pass it to Claude or ChatGPT and tell it to build the app."** That's it. You will learn in the process, but let's not assume that you know more than these systems in many areas. We may not.

**Johan**: [13:16] Thank you. I maybe have one question. When you said the user providing feedback of the product, does that mean that I should add a feedback section inside?

**Anand**: [13:30] That's a possibility. But I would do that if you have people who would be passing feedback without you being there. Right now you have just few users and you can ask them for feedback at any point. So is it worth adding that feature? Think about it. **It's very easy to add any feature. It's easy to create a version with feedback, without feedback. It's easy to do anything. That being the case, there are a thousand things that you can do.** You could literally ask it: "Think of a thousand features and add it." But **which of those will be useful becomes an important question.** And that's a big part of what today and tomorrow is really about. We can build anything, we can add anything, we can remove anything. **What should we add comes from feedback from real usage.**

**Anand**: [14:33] I'll try to implement and iterate. Thank you.

**Anand**: [14:35] Another thing that everyone should keep in mind is: look, everybody has ChatGPT, everybody has Claude. You can build an application; they can build an application too, just as easily. **Why do we even need an application?**

**Anand**: [14:52] So for instance, what would it take to just ask Claude or ChatGPT the question: "What should I trade on next?" What if you just told it: "Look, maintain a ledger of my portfolio on an Excel sheet somewhere. Just keep it and go fetch from whatever site you think is best all the information." In other words, we are so used to specialized software doing things that **we forget there's one general-purpose software now, or a few like Claude and ChatGPT, that can do the work of most software.**

**Anand**: [15:35] If I wanted news summarized, then I would just go to ChatGPT and say: "Summarize the news for me," and it gives me the same summary. **Why do I need an app for that? Which is not to say that we shouldn't have an app, it is to say that we have a similar capability elsewhere and we need to think about why _this_ is different and useful.** Great. Who'd like to go next?

**Johan**: [16:04] Thank you. I'll stop sharing.

**Christopher**: [16:31] Hello. I'd like to go next but unfortunately my shareable link is not working.

**Anand**: [16:41] Okay. Let me take a look.

**Christopher**: [16:43] Yeah, I think there's something wrong with my settings for my Claude. My artifact just can't be shared.

**KK**: [16:56] Sorry, I think yesterday I had this same issue. I'm interjecting, but I think there's a settings that you can toggle for it then you can...

**Christopher**: [17:05] Yeah, I tried toggling but it just doesn't allow me to toggle.

**Anand**: [17:12] Maybe you can share your screen and see if we can debug that.

**Christopher**: [17:16] Sure. Okay. Hold on. Give me a moment.

[Pause 17:21 - 18:05]

**Christopher**: [18:06] Can you see my screen?

**Anand**: [18:07] I can see your screen.

**Christopher**: [18:09] Yeah, so basically you usually can like share it here, right? But this is grayed out. And then after that if you go to my settings here, then you go to privacy, then you go to "Share Artifacts," right? So this is the one I built. Then I try making it public... it just can't be made public for some reason.

**Johan**: [18:41] Sorry, if I may interject. My problem was because my product allowed for downloadable files and right now the terms and conditions of Claude doesn't allow a product with downloadable files to be able to be used by other people other than the creator. So I changed it to "Export" instantly instead of "Download." I told Claude the problem and it fixed it for me.

**Christopher**: [19:20] Which I think is... I think for my application, right... so **my application is basically like an AI VC.** So a VC is like a Venture Capitalist, right? So basically before a certain—maybe you're like a founder of a company, right? Then you want to pitch to a VC. So basically **this app helps you to refine your pitch.** I think it uses Claude itself to evaluate like whatever you... let's say you take a seat, then all these like questions...

---

**Christopher**: [20:00] ...all these questions and answers are using Claude itself to evaluate it. So, I think that is the reason why I cannot share it. But yeah, like even if I try to, let's say, go to a separate website like Vercel or Droppify or something, it just exports the HTML; it doesn't export the full functionality of the app.

**Anand**: [21:21] Got you. And how would we solve that?

**Christopher**: [21:39] I have no clue. I mean I can ask, but I did ask and they gave me several options here which I don't know what to choose. As in like I chose the first... how should I say... like I tried to see if the settings can work, but then it doesn't work. Then after that I tried with the HTML, the index. So basically just a bare-bones HTML file, but then it doesn't function like it's supposed to.

**Anand**: [21:51] Yeah, which is a good place to be because now again you've discovered a problem that's worth solving. So, a few thoughts here. One, **you can run experiments in parallel. The way you could do that is by branching from the chat.** Can you just hover over any of the past chats and see... okay, no wait, hold on. Sorry, Claude doesn't allow branching, does it? Can you hover over any of the past chats that you have from just a little bit further up? Scroll up a little bit. Keep scrolling down. Sorry. Okay, yeah, perfect, you do have a fork from here. **So, pick any point that you want and you can have different conversations on—maybe try approach A, maybe try approach B, approach C. That's thought number one.**

**Anand**: [22:22] Number two, what you can do is **try things in parallel. You can say: "Run experiment A." While that's running, run experiment B. And while that's running, run experiment C.**

**Christopher**: [22:35] Okay. Yes, that makes sense.

**Anand**: [22:37] And second, you could also **tell it to take control of your browser. And if it needs to log into Netlify or Vercel and create accounts or whatever, tell it to go ahead.** Right now you're trying to understand what it's doing. Get out of the way. Let it do what it needs to do. You are just giving it instructions rather than trying to follow what it's saying and "Oh, should I do this? Should I do that?" Try to avoid having to make decisions. Let it run. You're watching it anyway, so if it does anything that you don't want it to do, you can always stop it.

**Christopher**: [23:22] Okay.

**Anand**: [23:24] Another possibility that you can think of is **ask it: "If you had to rebuild this from scratch so that I can easily deploy it, how would you do it?" That's another kind of experiment.** Because now it knows where it can deploy, the fact that you want it to be deployed.

**Anand**: [23:52] Fourth possibility: start from scratch with one extra statement saying: **"I want to be able to share this," and maybe you can tell it where to share it at—Vercel could be an example—or you could say, "I will run it on my machine as a separate server,"** however you want to deploy it. But those were ideas for deployment. And yeah, to show it to people, because that's when you get more feedback, capture the feedback, etc. Sorry, go on Christopher, you were going to say something?

**Christopher**: [24:31] Uh, nothing.

**Anand**: [24:32] So one thing to take away from this—and I suspect the others will too—actually two things that you might want to take away is: **One, if you have AI being used as part of the application itself, that lends itself to more powerful use cases.** But the questions then become: "How and where do I deploy it? What if it costs too much? Who should pay for it? How do we handle that?" Those sorts of questions do come up and it's worth tackling those because that's when you'll get a sense of a few design decisions. You'll say: "Maybe I'll let them use a cheaper model. Maybe I'll say each user should log in and they get a dollar a week," something that's within your limits. That sort of thing is always a possibility.

**Anand**: [25:32] **The second thing that makes it interesting is the deployment challenge. How do you deploy a more complicated application where people do have to log in, we do have to track what they're doing, we do have to store an API key maybe?** Those sorts of things become interesting. So push the bounds. **One thing that I definitely like to hear when we catch up tomorrow is: "What are the failures? What did you try that didn't work?"** As many of them as you can, because again, collectively we will learn from that. Of course, "What feedback did you hear?" is an important question, but I would also add, "Where did it fail?" Another thing—even if you eventually succeeded—is: "Where did it do something wrong? Where did you have to step in? Why did you have to tell the agent to do something differently?" And in this case, the deployment was one such scenario where Christopher, you said, "Do this," and it couldn't. So failure can lead to that, but it may also be it did something that I didn't want it to do. That's useful too. Keep a note of these and we'll discuss some of those tomorrow.

**Christopher**: [26:54] Okay. Okay, thank you.

**Anand**: [26:57] Great. I think we have one more student. Sorry, I don't have your name on the meeting, but it's g2527083. Did you have anything you wanted to share or ask?

**Kosei**: [27:37] I'm just sharing real quick.

**Anand**: [27:41] Perfect. I can see the screen. That is interesting. Ooh, goals and where they were shot from. Sorry, what's your name, please?

**Kosei**: [28:41] Oh, sorry. My name is Kosei.

**Anand**: [28:44] Kosei. Kosei, feel free to speak in Japanese. I'll get the translator to sort it out.

**Kosei**: [29:02] 僕はこのサイトを作りました。で、まずこのサイトは何を表しているかというと、イングランドのプロサッカーリーグ、プレミアリーグにおけるシュートの位置とゴールの可能性の関係を示したものです。対象は2025年、2026年シーズンで、プレミアリーグに所属する全20チームおよび20チームに所属するすべての選手を対象としています。下の図はシュートの位置と、そこから打ったシュートのうち、何本の点が決まったかを示したものです。
[**Translation**: I created this site. First of all, to explain what this site represents: it shows the relationship between shot positions and the probability of goals in the English Premier League. The target is the 2025-2026 season, covering all 20 teams in the Premier League and all players belonging to those teams. The diagram below shows the positions of the shots and how many of the shots taken from there resulted in goals.]

**Anand**: [31:11] This is fascinating. **I guess a coach would find this really useful. Now, maybe you could put yourself in the shoes of a coach, or better yet, can you show it to a coach and ask them how this is useful, how would they use it?** Was that clear? Okay, I'll try to explain him. Sorry. No, no, actually let's try something different. I'm going to try with ChatGPT and say: "Translate whatever I say into Japanese, please."

**ChatGPT (Voice)**: [32:15] Got it. I'll translate your speech into Japanese in real time.

**Anand**: [32:21] **The application would be really useful to a coach. You might want to show this to a real coach and see if they can use it in a practical way. Do you know any coaches that you could show it to?**

**ChatGPT (Voice)**: [32:38] アプリケーションはコーチにとって本当に役立つと思います。これを本物のコーチに見せて、実際に実用的に使えるかどうかを確かめてみてもいいかもしれません。それを見せられるコーチを誰か知っていますか？
[**Translation**: I think the application would be really useful for a coach. It might be a good idea to show this to a real coach and see if they can actually use it practically. Do you know any coach you could show it to?]

**Kosei**: [33:00] えっと、僕自身が紹介できる実在のコーチはいないんだ。ただ、もし実際に試すなら、サッカーやバスケのコーチ、学校の部活コーチ、パーソナルトレーナー、スポーツアナリストなんかが良さそう。
[**Translation**: Um, I don't personally know any real coaches I can introduce. However, if I were to actually try it out, it seems like soccer or basketball coaches, school club coaches, personal trainers, or sports analysts would be good.]

**Anand**: [33:22] Oh, we didn't hear well. Oh, okay, it's okay. Yeah, I'll explain in Japanese.
[Anand speaks Japanese to Takayuki/Kosei]
**Anand**: [33:31] コーゼイ君が、まあ、そのデータ集めただけや。なんで、その、どういう使い方が、例えばチームのコーチに見せて、それが役に立つかとか、その、なんか目的っていうか、それがどういう風に役に立つかみたいなのをちょっと説明... ああ、OK、説明っていうか、その機能を付けて... ああ、なるほど。
[**Translation**: Kosei-kun just gathered the data. So, for example, show it to a team coach and see if it's useful, and explain the purpose or how it would be useful... Oh, OK, not explain, but add those features... Ah, I see.]

**Anand**: [33:54] Sorry, was that a "Yes, there are some coaches you could show it to?"

**Takayuki**: [34:05] I think he doesn't think that kind of stuff yet. I think this is just statistics.

**Anand**: [34:14] Absolutely. But who is it useful for?

**Takayuki**: [34:18] Okay, basically coach and team and some kind of director of the team.

**Anand**: [34:26] And I would suggest showing it to a few such people. Or if not, **just mentally wearing the hat of a coach: "How would we use it?" and make actual suggestions. Maybe add a few people to the team, tell these people to play from here, some people are stronger in certain areas, make passes on that side. In other words, once it starts becoming useful, that's when it turns into a product.** But this is obviously very impressive. Now the next step is finding out the use. I'll share this in any case as part of the meeting notes.

**Anand**: [35:21] I think that covers all the students that are on the chat. Just a quick check then: any questions from anyone?

**Anand**: [35:43] Okay, I will then summarize the minutes of this discussion. But effectively for tomorrow, **what I'm hoping is you would have definitely shown this to a few people, recorded the feedback, and will be sharing the feedback. You will also be sharing where you needed to tell it to do something different or extra and why.** What we'll do is tomorrow pivot to the next stage, which is: **"How do we know the app is working? How do we verify it? How do we make it useful? How do we go about deploying it? How do we share it? How do we protect it?"** All of those sorts of things. As before, if you have any questions just send an email to askai@straive.net. I'll put that in the chat as well.

**KK**: [36:39] Wait, wait, one... I just thought because you said AI, right? **Can we pass the app to like ChatGPT or something then get _that_ AI to evaluate it?**

**Anand**: [36:51] **Absolutely! And you could ask it to play the role of different personas as well. You could say: "Look, for instance, I want you to act as a football coach; review this app. I want you to act as a football enthusiast; review this app."** I would ask it what other persona and it may say, "Oh, maybe we could be a media company or a sports channel," and review this like a sports channel. You could almost do a survey, you could have it do a business strategy. It may not be right, it may not be good. The smarter models may do a better job. I'd say Astra or Fable will probably do the best job if you have enough credits, but it may not be perfect. So human feedback is going to be important. But yes, they are pretty good at this sort of review as well.

**Anand**: [37:53] Cool. Thanks for joining in, folks. I'll see you online tomorrow. Bye.

**KK**: [37:59] Thank you. Bye.

**Kosei**: [38:01] Thank you.

**Christopher**: [38:02] Thank you. See you tomorrow. Bye-bye.