Streamline your end-to-end developer workflow with AI | ODSP908

0:00WILL WANG: Hi, everyone.

0:00Today we're going to talk about how

0:01to streamline your end-to-end developer workflow with AI.

0:05My name is Will.

0:06I'm the founder of Clova, where we build an AI agent

0:09that allows different brands and businesses to be able

0:11to quickly generate on-brand assets

0:13and iterate through them as well.

0:15We're also backed by the Founders Inc.

0:16Accelerator here in San Francisco within Fort Mason.

0:19And, in my free time, I enjoy making YouTube videos

0:22where I document my startup journey

0:23with over 18,000 subscribers on the channel.

0:26And I just share about the ups and lows of building a startup

0:29and also share about how I build software.

0:31And today we're going to cover the four phases

0:33of shipping a product end to end.

0:35In the first phase, we're going to talk about how to come

0:36up with startup ideas.

0:38In the second phase, we're going to talk about how do we build

0:40with Copilot and VS Code.

0:42In the third phase, we're going to talk about how

0:44to effectively communicate with our team so we're able

0:46to move fast while building quality software.

0:49And, lastly, once we have our software built,

0:51we're going to talk about how do we go to market

0:53to get real-life users.

0:54To come up with the idea, the number one thing I suggest is

0:57to build for a problem you have for yourself.

0:59The first reason is that, if you already run

1:01into the problem yourself,

1:03most likely you already understand it

1:04on a very deep level.

1:06And the second reason is that, for a problem you already have

1:08for yourself, you will also be using your own software

1:11to resolve those issues, right?

1:12So you're able to go through rapid iteration loops.

1:14The second thing is to be a domain expert in a field

1:17that you're trying to solve the problem for.

1:19Again, that also helps with understanding the problem.

1:22The second thing is that you're able

1:23to most likely know other people in the industry already,

1:27so you're able to find early adopters very quickly

1:29through your network.

1:31And a lot of times people think, when you come

1:33up with a startup idea, you want to sit down and think

1:35about what good ideas there is.

1:37But, from my own experiences, a lot of times all the ideas,

1:41where they come from is when I'm not actually actively trying

1:45to think of different ideas to build.

1:47For example, let's say if I'm browsing the internet,

1:50and what happens is that you always want

1:52to catch the ideas the moment they come

1:53up because they can very quickly go away.

1:55So something that has been helpful is

1:57that I'm using the Logitech MX Master 4 over here is

2:00that they actually have an action ring here.

2:02So over here I configure it over here to open up a new notepad.

2:05So I have a documentation where I'm able

2:08to quickly write out different ideas.

2:10Another thing that's also helpful is that a lot

2:12of my ideas come from when I'm actually actively reading

2:15different articles, consuming new information.

2:17So I also use the screenshot feature very often here

2:20to quickly screenshot the parts, so I can also save

2:22that into my notes so I have it for reference.

2:25And, lastly, something that's also helpful is just to play

2:27and pause my music, honestly.

2:29Sometimes that puts me in a different mode when I'm thinking

2:31of different ideas or also just when I'm, you know,

2:33working on different coding projects.

2:35The last method I suggest is to run user interviews

2:38about their day to day.

2:39So, for these, I tend to focus on questions

2:41like what do you spend the most time on, what is your day

2:44to day like, and what does your organization have budget for?

2:48Because those things are very deterministic, right?

2:50They're not very opinionated.

2:51Since they're just describing what is happening,

2:54they won't usually have a bias

2:55when they're answering the question.

2:56So, moving on to the building phase,

2:58so the three major use cases I've been using AI for,

3:01number one, will be obviously writing code with GitHub Copilot

3:04but also writing a lot more documentation,

3:06a lot more tests as well.

3:08For the documentation and test part, I have already tried a lot

3:11of different prompts and also iterated a lot to be able

3:14to consistently get the results I want, right?

3:17However, to be able to, you know, quickly find these prompts

3:19or copy and paste them in, I'm definitely not going

3:21to rewrite them in the GitHub Copilot Chat.

3:24I actually already configured my keys

3:26on the Logitech Creative MX Keypad.

3:28So each of these keys, when I'm in VS Code,

3:31actually is a different prompt.

3:32So I'm able to quickly paste in these prompts

3:34and execute these tasks without having to context-switch and try

3:37to find where these prompts are stored

3:39or rewrite them from scratch.

3:41And something that has also been very, very helpful is

3:45that because for each

3:46of the features I'm shipping I actually start a new AI session,

3:50right, because obviously with AI,

3:52a limitation is the context window.

3:54So I don't want to have, like, a feature that is not relevant

3:56to the next feature I'm shipping within there.

3:58But you're able to pass down the information very effectively.

4:01After I ship a feature, or after each session,

4:04I write down a very detailed AI doc or also for myself to read

4:08and for the next AI session to be able to catch up on

4:10within that feature, within that folder as well.

4:12A new thing that has also been happening with my team is that,

4:15in the past, we're always kind of debating a lot, right,

4:18about what features to ship, what experiments to run.

4:21But because of how much faster we have

4:24to ship software these days, instead of debating,

4:26we just try multiple features and go to production

4:29with it because, at the end of the day,

4:30we want to let the market decide and see

4:33where the pull is coming from.

4:34So, moving on to Phase 3, we're going to talk

4:36about how communication has also really changed with AI

4:39for different teams while shipping software.

4:41So, in the past, a lot of times I would just get on the call

4:44and explain to my team what I think is important,

4:46what the milestones are trying to hit,

4:48or what other software issues we're running into.

4:50However, with AI now, we're able to write

4:53down so much more documentation at a quick pace.

4:56I actually tell the AI to write

4:57down a very detailed HTML doc that's formatted very nicely.

5:00And I present that to my team because, when I'm on the call

5:03with them, now, they're able to really stay aligned more.

5:06And I also noticed they're able

5:07to digest the information a lot better,

5:09since there's something visual for them to follow along with.

5:11And I also obviously share those documentations with them

5:14after the call, so they also have that for reference.

5:16But, also, while showing documentation on the call,

5:20I also use Microsoft Teams for my calls.

5:22So, in Microsoft Teams already, it takes the transcript;

5:25and obviously the AI also writes down good documentation

5:29and different action items for the team to follow as well.

5:31Detailed documentation has always been super helpful

5:33for the code base and also on the business side, as well,

5:36but they just take so long to do.

5:37And the cool thing is that a lot of times you're able

5:39to get consistent behavior from the AI

5:41if you're using the prompt that you have already tested.

5:44So I actually also have the prompt saved

5:46on my Logitech MX Creative Keypad right here.

5:48So I'm going to quickly access that to write

5:51down these documentations as well.

5:52And, moving on, we're going to talk

5:53about once you have your app shipped, right,

5:55how do you go to market?

5:56How do you find users?

5:57So usually what I suggest is always to go

6:00where your users are gathering.

6:02And, for different products, there's different places

6:06that users gather, right?

6:07So a lot of the audience watching right now are

6:09probably developers.

6:10So, for the dev community,

6:11usually what I suggest will be Twitter.

6:14Also, I suggest Product Hunt, since a lot of folks

6:16on Product Hunt are looking at different products to try.

6:18And a common way that people have also been doing is

6:20to create an open-source project, right,

6:22since an open-source project, if you build something cool,

6:24a lot of times people contribute to it; and they're very invested

6:28when they see cool software that's open source.

6:30And through GitHub and GitHub Stars, you're able to get a lot

6:33of visibility through that.

6:34For my own personal software that I'm building

6:36with Clova right now, I'm actually doing a lot

6:38of the marketing on Instagram,

6:40since a lot of my users gather there

6:42and through LinkedIn as well.

6:43And something that also has been helpful from my own experience

6:47and other folks that I see ship software is also building

6:50in public.

6:51So I do that with YouTube, right, where I kind

6:53of just document how I'm building.

6:54And people get invested in your journey, and they end

6:56up trying your software.

6:57And, once you're able to get users there, you want to be able

6:59to reach out to them through email or whatever platform it is

7:02since you want to get on these calls with these users

7:04to get feedback, right, to know how to build a relationship

7:07and obviously to know how to improve

7:09and iterate on your software.

7:11So a lot of times these messages are also very consistent.

7:14I have a couple messages already written up.

7:16So I actually already also have those saved

7:18on my Logitech MX Creative Keypad over here as well.

7:21So I'm able to, you know, simply just change a first name

7:23and then send out those messages very, very quickly as well.

7:26But the thing is building is still the skill.

7:28But the leverage now is that AI lets us build so much faster,

7:31so we can put up more shots and test out way more ideas.

7:33And, after I put it out there, we can see the signal

7:36from the market on if it's an idea we want

7:38to continue working on or not.

7:39But on the flip side, too, something I have been noticing

7:42for myself and others is

7:43that a lot more personalized software has been, you know,

7:46coming to life, as well, because in the past it's hard

7:49to justify the time and energy

7:50to build those personalized software, right?

7:52But now with AI you're able to bring those ideas to life

7:54so much quicker, which has been super, super cool.

7:57And I'm excited to see what you ship next.

7:59If you do, please let me know.

8:01I'm always excited to check out new projects as well.

8:05And I hope you learned something from this presentation,

8:07and thank you for the time.