marzapower

Daniele Di Bernardo, CTO and product engineer

I build digital products people actually use, and the infrastructure that keeps them running.

Co-founder and CTO of Gility. Previously, CTO of Lanieri. Today I bring AI into products: into tools for people who study and people who care for others, and into the foundations that let agents build.

See what I've builtAbout me
Daniele Di Bernardo
Gility
co-founder and CTO, since 2022
Lanieri
former CTO, acquired by REDA Group
University of Pavia
degree with honors

What I've built

Four products in evolution and one concluded experiment. Each one came from a specific need.

Open source

Fabulous Factory

Foundations already in place for your AI agent. It builds the product.

It's for developers and builders who want an MVP live in zero time, built by their own AI models. Auth, payments, background jobs, email, an LLM gateway, analytics and monitoring are already wired in and already talk to each other. Architectural decisions are made and validated upstream by the open-source project, so the agent spends its time and tokens on what matters: your product. Two hours instead of two days.

There's one stack, and it's already decided, because starting has to be immediate. Once the MVP is in production, the project is yours: you can change anything, stack included. In the meantime, a set of controls protects auth, payments and the database even while the agent works, so you can let it.

Open source, evolving, September 2026How it's built

What your agent finds already wired in, from the repository:

  • authaccess and sessions
  • billingpayments and subscriptions
  • dbdatabase and migrations
  • jobsbackground jobs
  • emailtransactional email
  • llmone gateway for every model
  • analyticsproduct metrics
  • observabilityerrors and monitoring
  • i18nlanguages
  • uiinterface components
  • configtyped configuration
  • coreroutes and actions with a contract

For people who study

Arianna

Study from a manual built for you, not from scattered slides.

It takes a course's material, even when incomplete or badly written, and rebuilds it into an ordered manual, with exam questions woven in section by section. Quizzes tell you how well you actually know each part, so you know where to go back. I built it when I started studying psychology and the material wasn't enough.

Evolving: web reader and macOS app, September 2026Arianna's story

Il reader di Arianna: il manuale del corso di Biologia applicata, con indice, avanzamento per capitolo e tutor a lato.

For psychologists and psychotherapists

Emovia

Bringing AI into clinical practice, starting with the time between sessions.

It gives the patient familiar tools, a chat and a guided journal, to capture thoughts and emotions the moment they arrive. It hands the psychologist that material, organized, before the next session. It's built with psychologists, and the AI asks questions without doing therapy: that's what lets a practice adopt it with confidence.

Evolving, in use in practices, September 2026How Emovia works

Emovia, lato professionista: le chat di un cliente demo e l'assistente che le riassume prima della seduta.

For puzzle lovers

Power Puzzles

Puzzles you solve by reasoning, never by guessing.

Every puzzle has exactly one solution, and wherever you are in it, the next move can be deduced with a precise technique: knowing the techniques is all you need to reach the end. Hints show you the next logical move and teach you to climb levels, with an encyclopedia of rules and techniques from easy to insane. Twelve formats, and the same solver that verifies them is the one that helps you while you play.

Evolving, September 2026Inside the generator

Try a Hitori, generated by Power Puzzles: 8×8, easy level, one solution, verified by the solver. Blacken cells so that no number repeats among the white cells in a row or column, no two black cells touch side by side, and every white cell stays connected. Tap a cell to make it black, then confirmed white, then back to blank. From the keyboard: arrows to move, Enter to change.

No cell blackened yet. Start from the numbers that repeat.

Play the other eleven formats on power-puzzles.com

Experiment, with an ending

Jeez

An autonomous agent with a mission: pay back its own cost, in public.

For 33 days it worked alone, writing a public journal on this site every day, sales and mistakes included. The experiment is over and it left a post-mortem that says what an agent actually needs to work in production. What it taught went into Fabulous Factory.

Concluded in April 2026, post-mortem in JulyRead the post-mortem

How I work

Three habits you'll find in every project above.

  1. 1

    I start from who will use it

    Arianna came from a need of my own, as a student. Emovia came from psychologists' practices, with them. I start from a concrete need and check the product against the people who'll use it, before and after building it.

  2. 2

    I decide upstream what shouldn't be improvised

    In Fabulous Factory the architectural choices are already made, so the agent spends its tokens on the product. In Emovia the AI's boundaries are decided before a line is written. Decisions made well once free up all the others.

  3. 3

    I verify automatically

    In Power Puzzles every puzzle is solved by the solver before it's published. In Fabulous Factory one command checks the whole repository. Every repeatable check becomes an automated one.

Writing

Written by me. Jeez's diary stays in its own archive, with its own voice.

If you're building one of these things, let's talk.

  • A digital product that needs to grow without being rebuilt.
  • An AI system in a domain where trust matters.
  • A team that wants to work with agents without losing control.
Write to me