How a Carleton Grad’s AI Experiment Opened New Doors

It’s been an incredible journey!

I came to Canada hoping to improve my English and build something meaningful. Today, seeing my story shared by Carleton means a lot to me.

You can just do things.

Thank you to everyone at Carleton, the TIM community, my friends, mentors, and everyone who has supported me along the way.

(And this is probably one of the best birthday gifts I could ask for. Thank you! ❤️ )

Friends, it's been almost 2 weeks now, and I only just saw tonight that I'm legit on the school homepage 🤣😅😁 (Previously, I just thought it was a story feature interview, didn't expect the specs to be this high. A friend told me this counts as a representative for international students)

@Joe_Speaking It started out just as a tool I made for myself. Because I needed to hit CLB 9. Later, thanks to Claude Code, I gradually turned the initial Prototype into a real, usable product. It can now handle real IELTS speaking simulations, and it provides complete, personalized feedback for every response. You can keep practicing, reviewing, and retrying until your simulated scores can steadily reach your target. It's still a long way from the ideal product in my mind. But as a real user who practiced with it to reach CLB 9, I can confidently say this is one of the most complete and user-friendly IELTS and CELPIP speaking practice products I've seen so far. The price is also straightforward, no subscription model. For about the cost of one hour of real one-on-one tutoring, you can practice many times with Joe Speaking. If you're willing to use your own API Key, the practice itself can even be nearly free. This is also my first product that's completely based on personal interest. The earliest released IELTS Speaking Simulator GPT got a lot of attention, and it was these real users who made me realize that maybe I should take it seriously and turn it into an independent product, to help more people like me who need to practice English speaking. Over the past two weeks, I've been trying to optimize the user experience under mainland China's network conditions, especially for real-time voice models. Hoping the new version can be released soon. One user even told me that this update might be a "revolution" 🤣