rethinking the ENT, solo and with AI

How it all started

The ENT is the university’s online portal, and my whole student life runs through it: my courses, my grades, my messages. The problem is that, like most of the digital tools schools hand you by default, it’s particularly frustrating to use: slow, laggy and honestly not very practical. Unlike Pronote, the single centralized platform you use from middle school through high school, higher education makes keeping track of your own studies a real challenge: services scattered all over the place, sometimes several logins, and nothing connecting the data.

The idea of bringing some order to this system had been on my mind for a while, given how much time I lost just getting to a simple piece of course material.

One Sunday morning, with an urgent assignment due that evening, I had to try again and again, between stuck redirects, expired sessions and loading screens that felt endless. That’s when I told myself I had to do something. So I started building Tafac, one single app that brings together everything students need, for every university.

It’s an ambitious project, sure, but with the help of AI and my UX and UI skills, I got a convincing first version off the ground, already used by a good part of the IUT de Lannion (my technical university college).

Solo from A to Z

On Papillon, I design and developers implement. On Tafac, there’s nobody else, and this is the project that let me discover a whole workflow with AI.

No more handoff: I don’t need perfect mockups or specs for someone else. I make the mockups that really matter to get the art direction across to the AI, and it implements the rest.

In the mockup, every widget had its own error state: “unavailable”, with a button to try again. Using the app every day, I realized the university’s services go down a lot: some mornings, half the dashboard was showing errors and Tafac looked broken when it was actually the ENT that was. So I went the other way: a card that doesn’t load disappears, and comes back on its own as soon as the service does. That’s the kind of decision no mockup would ever have led me to.

The flip side is that nobody challenges my choices. The AI isn’t a student, and I’m in my own student bubble, alone with my project idea. So I ask around, talking to students who actually use Tafac to get relevant, specific feedback.

Coding with AI

I’m not a developer. I know the basics and can code in simple languages, but a project as complex as Tafac, which involves reverse engineering, was way beyond my skills. Luckily, it’s 2026, and today there are plenty of AI models that can do what I want, more or less well. And I won’t pretend otherwise: all of Tafac’s code was written with the help of AI. What interests me here is how the roles are split.

To hand my mockups over to the AI (here, Claude Code), I use the Figma integration for Claude: I send the link to my frames and Claude automatically pulls in the layers and turns them into code, which makes building and implementing my mockups down to the pixel a lot simpler.

Where the AI can trip up, at least the part I can see, is the UX. Sometimes the AI wants to create new sections of the app, but that often ends in a mess and a much worse user experience. As for the code, the backend especially, the mistakes or detours the AI makes are invisible to me, so I sometimes ask the backend developers I know for their opinion and ideas for improvement.

To build this app, I had to learn how to set up what’s called a “workflow”: deploying several agents on one big task, to split the work up and work efficiently. That system was new to me, since I was used to working with AI conversationally.

The trade-offs

Given how complex the French university system is, offering one solution for every university isn’t easy, even less so when only one person is on the project, AI or not.

So I had to make a few trade-offs, especially on scope, where I put my own university (Université de Rennes) first, to get something in front of me and my class as fast as possible. The system is still designed to be modular and take in other universities later.

Some features are also really complex to implement, starting with grades: in a BUT (the IUT bachelor’s degree), they’re organized by skills, with a fairly complex calculation system. So I chose to stick to the essentials for now: Tafac shows the latest grade you got and links to the official portal for the details. That protects the heart of the project, which is getting to all the different services faster.

Launch and first feedback

The first feedback was really positive: first picked up by my class, it won students over straight away. They found it simpler… and saw it saved them a lot of headaches.

That’s actually what pushed me to make it a real project rather than a side project, one where I can bring real value to students at the IUT.

Looking back

Tafac showed me that today, a designer can bring a real product to life, from idea to launch. The AI wrote the code, but it never decided for me what to build, or what worked.

The project is now open source: anyone can contribute and improve it their own way. To try it out, head to tafac.fr: a demo mode lets you look around the app without a university account.