A Field Guide to Interesting Problems

Gabriele Salvo
Eindhoven, The Netherlands
February 2026 — preprint, under continuous revision

Abstract

This document catalogues the ongoing investigations of one curious Italian relocated to the Netherlands. Topics span AI engineering, bionic prosthetics, haptic systems for healthcare, the aerodynamics of old Japanese motorbikes, and a hypothesis that the best ideas arrive while restoring something broken. A live multiplayer experiment is included for interested parties.

1. Introduction

Some people optimise for career paths. I optimise for interesting problems. This has led me from building predictive AI for prosthetic limbs to designing haptic devices for hospital patients, from teaching Python to teenagers to advising companies on the EU AI Act.

The common thread is curiosity. The secondary thread is stubbornness. Formally:

\[ \forall\, p \in \text{Problems} : \text{interesting}(p) \implies \text{investigate}(p) \]

I'm based in Eindhoven. I grew up in Italy, which mostly explains the hand gestures during code reviews.

2. Current Research Instruments

The following tools have proven useful across experiments:

Definition 1 (Interesting Problem). Any problem where the solution doesn't exist yet, or exists but could be remarkably better. See also: the reason I have too many browser tabs open.

3. Selected Experiments

3.1. Predictive Control for Bionic Prosthetics

Winner, PhD+ 2024 (CLabs × University of Pisa). The idea: make a prosthetic arm anticipate user intent before the signal fully arrives. Target latency: \(\approx 0\) ms. Think autocomplete, but for your body.

3.2. Immersive Haptics for Healthcare

Winner, ICMS × HaptonTech Master Challenge 2025. A modular haptics suite for bedridden patients — designed for reliability, low cost (35% cheaper than comparable kits), and the kind of comfort that's hard to quantify but easy to feel.

3.3. Teaching & Certification

150+ students across Save the Children, CoderDojo Pisa, Hack Your Future, Volta Institute. 90%+ completion rate. Certiport-certified instructor (AI & Python). 8+ certifications carried.

Theorem 1. The best way to understand something is to explain it to someone who has never seen it before.
Verified experimentally across 150+ instances. \(\square\)

3.4. AI Consulting & Sovereign Architecture

I help companies build AI systems that they actually own — no vendor lock-in, no accidental GDPR violations, no "the model is a black box and nobody knows where the data goes." For the full portfolio, see salvoaistrategy.eu.

4. Parallel Investigations (Off-Clock)

Not everything interesting requires a compiler.

Definition 2 (Barn Hypothesis). Given a sufficiently old barn and a collection of neglected Japanese motorbikes, the process of restoring both to their original splendour produces a state of deep satisfaction not replicable by software.

The long-term plan involves land, animals, a workshop, and a garage full of vintage engines waiting to run again. Until then, the research continues in smaller spaces.

Other active threads: martial arts (owning agility and strength, not just typing speed), strategic games (chess has been a constant — and yes, Yu-Gi-Oh still counts Some things you grow up with become part of how you think. Trap cards taught me more about timing than most textbooks.), and books — especially dystopian fiction and the oniric worlds of Murakami, where cats disappear and reality is negotiable.

Lemma 1. A restless body makes a restless mind. Training both is not optional.

5. The Experiment (Your Turn)

Every visitor to this site shares a single chessboard. Moves are stored in Redis and synchronised globally. It's a small proof that strangers can collaborate on something — even if the result is chaos.

♟ Play Chess Against the World

6. Open Questions

7. Conclusion

The work continues. The barn is still theoretical. The motorbikes exist only as bookmarks. But the interesting problems keep arriving, and that's the part that matters.

Q.E.D.