Around three months ago, I published my first repository on GitHub. This was something that I would never have imagined could have happened. Before then, I only knew of GitHub from my friends who were doing Computer Science; I knew it was some "geeky" thing and apparently involved a lot of contributions. But here I am, and what an interesting journey it has been so far. Today I’m putting the final finishing touches on a tool that searches companies & directors in New Zealand, and gives a better view of previous & current insolvency status than other credit-related tools (in my opinion). I’ve got at least 4 - 5 more tools in the pipeline. I’m building new tools, and enjoying it.
I write this to document how my own "vibe coding" journey started and a reflection on joining what is, in my opinion, the most exciting group to come out of legal tech: the LegalQuants (LQ).
I remember my first foray into generative AI back in 2022 when I got myself a ChatGPT subscription. At the time, I was very excited as to what it could do, assuming it would turn me into a superstar and do all my work for me. I was incredibly, incredibly wrong. I learned, rather embarrassingly, that AI hallucinates (a lot). I also discovered my first lesson about context windows and memory and how it simply couldn’t remember conversations from the previous day. I felt jaded and disappointed, and I cancelled my ChatGPT subscription after 2 months.
But AI has morphed into something, or probably more accurately, evolved into something far more sophisticated. When my work introduced Copilot, I asked to join it more "for the lols," but in return, I found that it was able to create genuinely interesting things. However, I would say that the use case and my expertise with it were still incredibly low. I was probably creating the equivalent of "work slop" most of the time, but there were a few gems. I was able to use AI to draft quite a few documents, including revised policies and recommendations for litigation work. But AI at this time was more of a chatbot or an assistant: I say something, it does something for me. Nothing more, nothing less.
It was only after I read an article about the concept of "vibe coding" that I found it interesting. I assumed it meant being able to code from scratch (which it kind of is), so I eagerly tried to use Copilot to vibe code a PPSR training app. I posted it in a previous Linkedin post and while I did create something, the outcome was not that great and it took far too long to do, But that little experiment meant a lot to me because it was the first time that I was able to use AI to not just generate a text output, but to actually create something, something tangible that I could show people and others could use. Was it good? No, it was pretty terrible, actually. But it was my terrible creation. I made it, I built it (of course with the help of AI). I am not naive enough to call myself a coder by any means, far far from it, but the point stands: AI was able to help me create something. I think the joy and pride of being able to build, create and enjoy its fruits is a common, universal human experience. If I were to get a bit theological, I think that is built into how we are designed; it helps us understand the distinction between God the Creator and us as His creation. But that’s a separate story for another time.
At this time, my thinking was that AI could help me create something, but that something probably wasn’t going to be very good. I thought of it as just a novel experiment, something maybe other people could figure out, but not me.
That is until I saw Jamie Tso’s post that went viral on LinkedIn, where he created a contract tabular review. He thought it was amazing. Many people thought it was amazing, and I thought it was amazing too. That led me on the journey to try and use Google AI Studio for the first time, and I created a contract review tool for laypeople to use. It was just for fun, but what I didn’t expect was, number one, how easy it was to create, and number two, how good it was. Suddenly, it opened whole new doors and potential opportunities, and that’s how I started on this building journey.
I write this post to maybe encourage other builders who are new in this journey. I am not writing from a place of experience; I’m writing from the place of a beginner, someone who is still learning lots and really enjoying the process. Below are some recommendations that I would give to someone new, and things I have to remind myself of every now and then.
First: Enjoy the build. The primary reason you want to build something is because you enjoy it. As mentioned before, there is great pride in seeing an idea that you may have thought far from possible turn into reality before your eyes and become something that you can digitally touch. It’s amazing. That joy should be the main impetus that propels and drives you, which then compounds into other experiences of joy. Suddenly, a problem that you have been trying to resolve has now been solved; miraculously, a process that has taken manual steps and many wasted hours can now be resolved in minutes or seconds.
Second: Embrace the MVP. It is good to be ambitious, ambition drives and fuels some of the world’s best ideas. But sometimes ambition can also kill an idea. If your product or app doesn’t look exactly like how you want it in your head, or if factors like security and deployment into production (especially in a corporate environment) scare you into even starting, don't be afraid of the MVP (Minimum Viable Product). Don't be afraid to create something that serves as a demo. After all, it is still your demo. You still built it. It is your proof of concept that you can then show and bring to other people to hopefully buy into your vision, or at the very least, help others see that you are passionate enough about the problem that you are trying to solve that you actually went ahead and created something.
Third: You will get better. The AI tools used to create these apps will become better and better, and so will you. Public perception of AI will always be polarising. There are some who deride "vibe-coded" apps and will say that you are not a "true coder" and that you actually didn't do any of the work. Fine, I’ll concede the first point: I’m not a true coder (but maybe you are). However, I don't think it's accurate to say that the AI tool is the one doing all the work. When I started my vibe coding experiments, I was literally being held by the hand by the AI. I had no idea what to do when an error occurred; I just asked it to help me explain and interpret it. I was just asking questions and following instructions.
But that eventually sharpened my skills. Now, even though I am far from an expert, I am able to identify what sort of issues the AI might come across and I am able to troubleshoot things, not because I’m smarter than the AI, but because I know the context of the issue, something that the AI may not have been aware of. I have gained confidence in how I structure my prompts, how I think about user design, and how I think about production-ready work in general. I now look at the console when something goes wrong, and can point the AI to the right API endpoint (because I’ve actually spent time reading the documentation). Don’t ever lose confidence or sight of the fact that you are also growing and improving as you build more.
Fourth: Just try & build something. Do it now. The investment costs of building these apps have never been so low. Apps like Replit, Lovable, Cursor, or even basic LLMs like ChatGPT, Claude, and Gemini offer some amazing tools to play with. Yes, there is a constant time investment, and with me having two kids, I’m not naive to the fact that time is a premium for most. But give it a try; you’ll be surprised at what you can build in a short span of time if you focus and think carefully about what you want to build. And if you’re really cheap skate (like me), Google AI Studio is free!
Finally, join a community. Despite all the talk about AI, it cannot replace human connection and belonging. Yes, you could simply ask your AI to teach you all the tricks and tips and guide you, but there’s something different from learning together with real people, doing the work and experiencing the same issues you have (I have other not so positive thoughts on AI….more on that next time!). You learn you’re not the only one trying to build something, and it’s a great feeling.
If you’re reading this and not a LQ member because you think you’re too new to AI or not good enough to be part of it, I strongly encourage you to apply anyway. Start building to show that you are someone actively thinking of issues and willing to try.
On the LegalQuants, I can truly say that this group of individuals are some of the most brilliant, talented, and most importantly, kind individuals I’ve been involved with. It’s a global group filled with lawyers from all walks of life, but they are all united under the same common goal: building something and developing skills that set them apart in our legal tradition. They aren't just selfish lawyers who aren't willing to share; they are incredibly generous with their time and very, very passionate about what they do. I am so grateful to have this group, and it has also helped me gain new ideas and sharpen the tools that I have built and will continue to build.
Every LQ member is amazing, but I want to give shoutouts to two individuals especially. The first is Yu-Chou (YC). This guy is a SuperDoc genius (he won’t tell you that, he’s very humble). When I had first asked a very basic question on the WhatsApp chat "How do I publish a Word add-in?" he private messaged me separately to give me some ideas and feedback. At the time, I had no idea what he was talking about, but I was able to research what he mentioned and was very appreciative that he reached out in the first place. A few months later, I then posted on Discord about my own redlining tool that I was developing. It was still in its infancy stage and honestly was not performing as well as it should (as to be expected with red-lining). Again, YC private messaged me as he was interested in learning more about what I was trying to build. I gave him my idea, he gave some feedback and direction, and overall it was just a very helpful and encouraging conversation. These are the type of people that you will meet in the LegalQuants: people who are genuinely interested in solving problems and more than happy to help you solve yours too.
Last but certainly not least, the man himself: Jamie. When I posted on LinkedIn about my contract review tool, admittedly Jamie was the last person that I would have expected to comment on it. The tool I made was actually quite basic, after all, you could just point your AI to a contract and it could probably do the same thing. Sure, mine might look a little bit fancier, but what’s the difference? Why would he be interested in it? A week later, he messaged me to ask if I would be interested in joining the WhatsApp group. I thought it would be quite interesting, so I said yes. I’m so glad that I did; if I had said no, I would have missed out on joining the LegalQuants. Sometimes I ask myself: why on earth did he invite me to join? The skill level and sophistication of both the lawyers and the tools they have built far surpass mine.
As I write this, LegalQuants is currently hosting another hackathon with Replit, where 50 lawyers will have one week to build an app and present it for judging. When I listen to some of the contestants share their vision for what they want to build, I can see the excitement but also some uncertainty in what they want to build. These are the same emotions and feelings that I definitely felt a few months ago and, admittedly, probably still feel today. If I saw someone build a tool and they tell themselves, "It's not that good," I would tell them to look past the quality of the tool itself. I would encourage them and tell them that deciding and making the effort to build something in the first place shows creativity, a willingness to try something new, a desire to not follow the status quo, and exhibits a very important trait: the willingness to create. The willingness to build something that they can call their own. Will it be better than the competition? Maybe not. Will it be something that is super scalable? Maybe. But the point is, it’s not about the tool that they create, but about the passion, the desire, and the skills of the creator. I don’t want to toot my own horn, but I have a feeling that maybe Jamie saw that in me. I wouldn’t know, I would have to ask, but if he did, I’m truly grateful, and that is something that I will not lose or take for granted.
So, to the Hackathon contestants: happy building! It will be fun, and you have the best support in the world helping you out.
To the ones who sacrifice their own personal time and sleep just to make something new.
To the ones who start panicking when the change they thought would bring a significant improvement starts crashing the whole app.
To the ones who sigh in frustration that their Claude code limits have already been reached.
To the ones who are sharing news, updates and repositories to learn and share something new.
To the ones who have successfully shipped their new tool, only to go back into it and add more features because nothing is ever truly finished.
Here’s to the builders.