Speaker Background

I'm Rick, Ricardo. I changed my identity to Rick recently because it's easier.

So, I was raised, I'm gonna put, yeah, I was basically raised with a computer in my hands. By the age of four, I was given an unrestricted computer, which can be both a genius idea or a terrible idea. And I hope that, yeah, my future is more on the first side.

So I love working with computers. I don't have a computer science background, but I just love building software, understanding how it works. That's why the back end of MindStone was kind of like very easy for me to ramp myself up to. And I also applied it in my professional life.

I've built on two software companies. One is Instify, the other one is High Radius. Instify InsurTech here in Amsterdam. them, and High Radius is a fintech operation company. Apollix is a Platinum partner for Salonis. Probably some of you guys have heard of Salonis, and BCG Platinum is where I did my first internship.

I currently work at Eleven Labs. Any one of you guys ever used Cloud? Okay, when you speak to it, the voice is given by Eleven Labs. So we basically give the voice to Cloud and many other voice applications.

I studied at the Erasmus School of Management and did my bachelor's in Maastricht.

So yeah you can connect if you want to scan the QR code and yeah I'll start with today's presentation. So before I go in yeah I can start like this.

Four Habits for Working Better with Claude Code

So I want to give credits where credits are due because I learned a lot about what to use and how to use AI by like from Nick Sarajev. Nick Starev is an incredible entrepreneur, a Canadian guy, and he does very easy to follow and understand videos on YouTube.

So you can scan this QR code to access the video that I based this presentation out of. Because he mentions 14 habits to use, or maybe more, I don't remember the exact number, to use when interacting with cloud code. And I selected the four that helped me the most and that resonated with me the most.

So without further ado. Yes, exactly.

Audience Check-In

These are the four. I'll need you guys to interact with me a little bit. So again, third time tonight, but this time with also a timestamp on it.

Who used the cloud code this week? All right, cool. Cool.

I will need you guys to keep your hands up for the ones that just had it up for if you had a long session that started giving you worse answers over time. Okay, so you're still keeping it up, perfect.

And keep it up if you also just put in, just fix it into your AI at some point in those moments. OK, cool.

1. Use Handoff Cards Instead of Compacting

Then this part is specifically for you guys, because when it comes to cloud coding specific, there is this capability of it to compact the conversation. Right now, it even happens automatically. That is not actually the way you're supposed to do it. Why?

Because your capability of the AI to use the context that you are constantly giving it to, which is usually the messages that you have with them, is very limited and you want to keep it rather clean. The slash compact command is not doing as good of a job as doing a handoff card instead.

that so what do you want to do you want to tell it summarize where we are what is done and the decision I need to make what's next and the open problems you ask it you receive a text and then you open a new session because if you were to just all the time compact the ongoing conversation

the yeah the AI ends up also taking in all the wrong decisions or the fixes that we're not supposed to be even kept in the same session so that's habit number one hand

2. Keep CLAUDE.md Current

off don't compact on the second one again if you guys can raise your hands if you have a Claude MD okay okay keep your hands up if you edited this month okay so I see that the same hands also edited that's very good and what's the the most, I mean, yeah, that's a little bit extra. You guys don't have your computers on, so it's okay.

Your CloudMD is basically the running, it's like the working capital of a company, if you think about it from the financial perspective. It's the running memory of whatever your Cloud instance is referring to all the time.

So keeping it up to date is incredibly important. You need to keep it current to what you're currently solving.

And to keep it good, because you don't want to always manually write it, because it's sometimes even more than 500 word document, you can ask in the same instance, how could you have done it faster with fewer tokens? And then update that one.

Now Cloud even updates it automatically, so you can just tell it to do it. So, yeah.

Sorry, you had a question? Yeah, you just need to check it yourself as well sometimes. That's the idea. Yeah, yeah, yeah, exactly. Because it might be that it does it automatically, exactly.

But in the example that Nick does in the video, there might be a core business function that you know as a human, right, that you have to change in the current interactions with the system, which cloud maybe doesn't know because it's some information that that it got very much down the line in the conversation or in the context that it had access to. So sort of like closing that gap of visibility can help you have more real conversations with it.

3. Define the Problem Before Asking for a Fix

Okay, and then last but not least, put your hands up if you let the model decide what broken means. Okay, that's very good practice. And, yeah, this is another habit that you should very much use when you want to fix stuff and do it the best way possible.

Instead of just saying fix it, which is something you guys already don't do, so that's very good. The main idea that you want to adopt is to first list all the problems and not write anything yet.

because it can be that it's instead of it being a bug it's a feature instead of it being something that might do it more efficient it loses for instance in the case of Nick brand identity because let's say that the AI if you were to say make it more efficient for me it would just see these as inefficient the opening is vague and maybe a business proposal it says the same thing twice the tone is too casual, and there's no example, but you want to keep it casual, right?

So again, trying to be more understanding on what the plan that a cloud wants to implement is helps you have better interactions with it.

4. Fan Out Cheap Models, Fan In Expensive Ones

All right, the last habit, which I believe is also the most interesting one,

is hands up if you ever changed model during a chat like if you from high cool you changed okay and what and hands up if you changed your effort doing the chat as well okay cool I know I probably very much believe that you guys know when you want to do it but this for me for instance was rather new like to have that as a kind of a mindset in the back of my mind on when to change it and for

what type of tasks so fan out cheap fan in expensive is the habit of having your question coming in right and then expecting the AI to read a lot of documents let that do by cheap models in parallel and then when it comes to the the decision, you want to, let's say, filter out from all that gathered reading and knowledge, yeah, knowledge reading, a mega prompt, we can call it, and then let the strong model decide, build, and then consult you on the information retrieved. So that's also a pretty simple one to keep in mind.

Questions and Discussion

All right, I wanted to still keep the last three minutes more or less for any questions or discussions. So yeah, that's that's it from my side