Skip to content

About

Understanding a system by building one.

Black-and-white portrait of Lorenzo Roque Dal Fabbro
Based
Brighton, United Kingdom
Born
Lisbon, Portugal, 2004
Heritage
Italian and Portuguese
Writes in
English and Portuguese
Studying
Politics & Philosophy, University of Sussex, 2024 – 2027

I’m Lorenzo. I study Politics and Philosophy at Sussex, and almost everything I make is a way of getting inside a system: building it, modelling it, or arguing about how it should work.

Sometimes that means simulations: economies where money is conserved to the cent, humans who get cold and build shelter, agents learning to play a strategy game. Sometimes it means software for my own life, which I treat as a system to be measured rather than guessed at. And sometimes it means essays and newspaper columns about the institutions everyone else lives inside.

I grew up in Lisbon, studied in Austin, and now live in Brighton. I started making things for an audience early: a Twitch affiliate at fourteen, about 204,000 TikTok followers by sixteen.

Five subsystems

The workThe writingGitHubSubstack

Education

  1. 2024 – 2027University of SussexBA Politics & Philosophy · Brighton
  2. 2022 – 2024St. Edward's UniversityEconomics, then Philosophy & Psychology · Austin, Texas
  3. 2018Imperial College LondonSummer course in machine learning and deep neural networks · London
  4. 2008 – 2022St Julian's SchoolIB Diploma, including Computer Science HL · Portugal

§ 01Trajectory

Compounding.

The same instinct at every stage, at a larger scale each time.

  1. AGE · 2018

    14

    Twitch affiliate

    and a summer course in machine learning at Imperial College London.

  2. AGE · 2020

    16

    204,000 followers

    on TikTok, grown from 6,000 in nine months.

  3. AGE · 2021

    17

    A political community

    built on Twitch around ideas, with its own website.

  4. AGE · 2025

    21

    34 essays, 7 columns

    in eight months, and the first simulations.

  5. AGE · 2026

    22

    Five subsystems

    PolyVision, Econ-Engine, Life Engine 3D and the Fabbro Systems apps.

Outside inputsNot chosen, but load-bearing

  1. Mar 2020

    Schools close

    Lockdown arrived mid-climb. The TikTok account went from 6,000 followers in September 2019 to 204,000 by June 2020.

    The content

  2. Dec 2022

    First use of ChatGPT

    Language models changed how fast I could iterate. Programming had come in waves; from March 2025 it stopped coming in waves.

    The work

  3. Apr 2023

    A journal, started

    Started the same week as the first essays on Substack: writing things down, privately and in public, at once.

    The writing

§ 02Beyond the screen

Two systems without a keyboard.

The same instinct, pointed at a body and a swing: log everything, change one variable at a time, and let the evidence decide what happens next.
Fig. 3.1Posing, 7 September 2026.

TrainingSince 23 Jan 2023

Progressive overload, logged.

Every set since January 2023 is in one ledger, now Heracles, the training app I built for it: load, reps and effort in, strength estimates out, with the uncertainty left visible. The programme is an upper/lower split, six days a week.

Open Heracles
sessions logged, 23 Jan 2023 to 1 Oct 2026
433
sessions logged, 23 Jan 2023 to 1 Oct 2026
working sets logged, plus 70 warm-ups
4,711
working sets logged, plus 70 warm-ups
training days a week, upper/lower alternating
6
training days a week, upper/lower alternating
years since the first logged session
3.7
years since the first logged session
Fig. 3.2The training ledger in four numbers.

The week

  1. UUpper
  2. LLower
  3. UUpper
  4. LLower
  5. UUpper
  6. LLower
  7. RRest

EvidenceTwo sets from 2 September 2026, both personal bests

Seated DB lateral raise2 Sep 2026
25 × 11kg × reps, per hand34.4e1RM, kg
Fig. 3.3Best lateral raise on record: +174% on the weakest logged session.
Seated alternating DB hammer curl2 Sep 2026
34 × 8kg × reps, per hand42.6e1RM, kg
Fig. 3.4Best alternating hammer curl on record: +478% on the weakest logged session.

Progression

From weakest to best.

On a log scale, the length of each line is the growth itself: the same distance means the same multiple, whether the bar started at five kilos or fifty.

Estimated 1RM, kg · log scaleWeakest sessionBest session
  1. Alternating DB hammer curl4.54 × 16 · 10 Feb 2023 · e1RM 7.434 × 8 · 2 Sep 2026 · e1RM 42.6+478.4%
  2. Wide barbell bent-over row27.22 × 8 · 7 Feb 2023 · e1RM 34.191.63 × 10 · 16 Dec 2023 · e1RM 122.2+257.9%
  3. Unilateral DB preacher hammer curl4.54 × 12 · 27 Jan 2023 · e1RM 6.520 × 5 · 7 Mar 2025 · e1RM 22.9+255.3%
  4. Seated DB overhead press, 60°9.07 × 10 · 23 Jan 2023 · e1RM 12.130 × 11 · 22 Dec 2025 · e1RM 41.3+241.4%
  5. Seated DB lateral raise10 × 8 · 30 Jan 2025 · e1RM 12.525 × 11 · 2 Sep 2026 · e1RM 34.4+174.3%
  6. 30° incline DB press13.61 × 8 · 30 Jan 2023 · e1RM 17.140 × 5 · 29 Apr 2025 · e1RM 45.8+168.5%
  7. Unilateral DB row22.68 × 8 · 27 Jan 2023 · e1RM 28.450 × 11 · 12 Feb 2025 · e1RM 68.8+141.8%
  8. 30° incline barbell press29.48 × 12 · 11 Oct 2023 · e1RM 41.980 × 5 · 17 Mar 2025 · e1RM 91.7+119.0%
  9. Unilateral DB overhead triceps extension8 × 7 · 29 Jan 2025 · e1RM 9.716 × 8 · 9 Dec 2025 · e1RM 20.1+106.1%
  10. Barbell press70 × 6 · 9 Oct 2025 · e1RM 82.7110 × 3 · 7 May 2026 · e1RM 118.7+43.7%
Fig. 3.5Ten free-weight movements, weakest logged session to best. Dumbbell loads are per hand; early sessions were logged in pounds.
Show as table
MovementWeakestBestChange
Alternating DB hammer curl4.54 × 16, 10 Feb 2023 (7.4)34 × 8, 2 Sep 2026 (42.6)+478.4%
Wide barbell bent-over row27.22 × 8, 7 Feb 2023 (34.1)91.63 × 10, 16 Dec 2023 (122.2)+257.9%
Unilateral DB preacher hammer curl4.54 × 12, 27 Jan 2023 (6.5)20 × 5, 7 Mar 2025 (22.9)+255.3%
Seated DB overhead press, 60°9.07 × 10, 23 Jan 2023 (12.1)30 × 11, 22 Dec 2025 (41.3)+241.4%
Seated DB lateral raise10 × 8, 30 Jan 2025 (12.5)25 × 11, 2 Sep 2026 (34.4)+174.3%
30° incline DB press13.61 × 8, 30 Jan 2023 (17.1)40 × 5, 29 Apr 2025 (45.8)+168.5%
Unilateral DB row22.68 × 8, 27 Jan 2023 (28.4)50 × 11, 12 Feb 2025 (68.8)+141.8%
30° incline barbell press29.48 × 12, 11 Oct 2023 (41.9)80 × 5, 17 Mar 2025 (91.7)+119.0%
Unilateral DB overhead triceps extension8 × 7, 29 Jan 2025 (9.7)16 × 8, 9 Dec 2025 (20.1)+106.1%
Barbell press70 × 6, 9 Oct 2025 (82.7)110 × 3, 7 May 2026 (118.7)+43.7%

The ledger

Ten standout sets.

An estimate, not a single truth: two formulas, both kept. The bar shows how far apart they land.

  1. 01Barbell press7 May 2026110 × 3118.7116.5–121.0
  2. 02Hip thrustNautilus Glute Drive8 Apr 2026240 × 7292.0288.0–296.0
  3. 03Chest-supported close-grip rowPrime Extreme Row7 Sep 2026140 × 8175.6173.8–177.3
  4. 04Close-grip lat pulldownHoist30 Sep 2026120 × 8150.5149.0–152.0
  5. 05Leg extensionMatrix G3-S711 Dec 2025118.7 × 10158.3both agree
  6. 06Seated leg curlNautilus Impact29 Sep 2026110 × 9142.2141.4–143.0
  7. 07Calf raise on leg pressAtlantis PW4121 Sep 2026140 × 13205.3200.7–210.0
  8. 08Seated DB lateral raise2 Sep 202625 × 11 /hand34.434.2–34.6
  9. 09Seated DB hammer curl30 Sep 202632.5 × 11 /hand44.744.4–45.0
  10. 10Wrist-cuff pushdownFFITECH Five Stations7 Aug 202685 × 7103.4102.0–104.8
Fig. 3.6Estimated one-rep max in kg: the midpoint of Brzycki and Epley, with both kept as the range. Bodyweight across these sets: 87.3–91.5 kg.

GolfSince Aug 2026

A swing, two months in.

I picked up a club in August 2026. Where lifting rewards force, golf rewards sequence, so I study it the same way: film it, slow it down, and work on one piece at a time. Scroll through the swing below.

7-iron · slow motiont 0.00 s

Phase 1/5

Address.

Set-up. Weight balanced, ball centred.

longest drive
222m
longest drive
fastest ball speed
247km/h
fastest ball speed
shots hit
1,443
shots hit
range sessions
19
range sessions
Fig. 3.7Range data, August 2026 to date. Swing filmed 22 Sep 2026.