Talk for PUG Torino - AI Tools

On April 17th, 2025, I had the pleasure of speaking at an event hosted by the Programmer User Group Torino, which I thank, at Toolbox Coworking, focused on AI tools for developers. In this article, I want to share the core of my talk, with practical examples and real-world use cases.

If you’re the founder of a startup or managing a small to medium business, and you need to build a digital MVP quickly, keep reading—you might discover a new way to speed up your project and drastically reduce development time and cost.

I began the talk by highlighting a key concept: the difference between predictive models and reasoning models.

• Predictive models classify, suggest, complete—perfect when you’re working with well-structured data.
• Reasoning models can write texts, make complex decisions, and—more and more often—write code and design interfaces.

Today, when I build MVPs, I rely on the latter. Why? Because they allow me to turn a text-based specification into working code quickly and effectively.

I also spoke about prompt engineering.

If building an MVP with AI still sounds like science fiction, the truth is that the real skill is not teaching AI to code—it’s learning how to communicate with it effectively.

A well-crafted prompt enables the model to:

• understand the business context,
• avoid mistakes and ambiguities,
• generate clean, structured code that’s ready to use or adapt.

In short, writing a good prompt today is like knowing how to code well yesterday.

During the evening, I demoed three tools I actually use when I need to validate an idea or launch a working prototype for a client.

Replit

With Replit, I started from a prompt like:

“I want you to create a to-do list web app with a basic UI and the ability to add, delete and mark tasks as completed.”

The result? A fully functional web app in just a few minutes, without setting up a local environment.

Replit also lets you instantly deploy the app online, making it perfect for testing with stakeholders or real users.

Best for: fast, testable MVPs—even without a backend.

Lovable

With Lovable, the prompt was something like:

“I want you to clone my company’s website and replicate both layout and navigation in a clean, maintainable codebase.”

Lovable quickly generated a semantic copy of the site, with clean code ready to be modified.

This is extremely useful for businesses looking to redesign their site or build an MVP inspired by an existing product.

Lovable opted to build the site using React, without leveraging Next.js’s SSR capabilities—which I don’t fully agree with. But migrating it was easy, and it’s exactly what I did to rebuild this website in just a few days.

Best for: fast landing pages, redesigns, and concept testing.

Cursor

The most interesting demo was with Cursor, an AI-native IDE that acts like a true software agent.

I asked:

“Please refactor the file tools.js, divide the functions by topic/area of math (e.g. algebra, geometry, calculus…) and create a new file for each area. Move the functions accordingly, delete the original file and commit the changes to GitHub.”

Cursor executed every step on its own:

• it identified the relevant math domains,
• created new files accordingly,
• deleted the original file,
• committed everything to GitHub.

Best for: production-ready code, clean architecture, and fast onboarding for complex codebases.

That was one of the most asked questions during the Q&A.

My answer is clear: no — but it will massively empower those who use it.

Today, developers need to learn how to orchestrate models, craft effective prompts, and validate AI output.

This calls for a new skillset—more strategic, more product-oriented.

And that’s exactly what I do when I help a startup or SME build their MVP: I integrate AI tools into the development workflow, or use AI in my own workflow to deliver a working MVP faster—without compromising quality.

Download the slides from the talk (available here).

“Artificial intelligence doesn’t just save time. It helps you build better products, faster—and listen to your users sooner.”

Let’s connect—maybe I can help build your next MVP.