My Stuff
Feeling like I might enjoy something that is not here? Please reach out!
Writeups and talks
Interpretability
-
The Engineer’s Interpretability Sequence by Stephen Casper (AI Alignment Forum, 2023-2025)
-
Mechanistic? by Naomi Saphra and Sarah Wiegreffe (BlackboxNLP, November 2024)
-
Downstream applications as validation of interpretability progress by Sam Marks (LessWrong, March 2025)
-
Assessing skeptical views of interpretability research by Chris Potts (Blog, August 2025)
-
A Pragmatic Vision for Interpretability by the GDM Interp team (AI Alignment Forum, December 2025)
-
Introspective Interpretability: a Definition, Motivation and Open Problems by Belinda Li (Blog, February 2026)
AI auditing
-
AI as systems, not just models by Andy Arditi (Blog, December 2024)
-
The fragile foundations of CoT monitoring by Chris Potts (Blog, July 2026)
AI safety and alignment
-
The ‘Breaking’ News: The OpenAI–Hugging Face Incident by Eric Wallace and Michael Dalton (Youtube, August 2026)
-
A retrospective of AI alignment by Richard Ngo (LessWrong, August 2026)
AI & philosophy
- What we talk to when we talk to language models by David Chalmers (PhilPapers, November 2025)
Personas and Simulation
-
Simulators by janus (LessWrong, September 2022)
-
The Rise of Parasitic AI by Adele Lopez (LessWrong, September 2025)
Research
-
You and Your Research by Richard Hamming (Gwern transcript, also video, 1986)
-
Principles of Effective Research by M. A. Nielsen (Essay, 2004)
-
How can we be good stewards of collaborative trust? by Chris Olah (Blog, May 2019)
-
How I Think About my Research Process by Neel Nanda (AI Alignment Forum, May 2025)
-
In Defense of Curiosity by David Bau (Blog, December 2025)1
- Ride the Speed Demon by Dirk Hovy (Blog, July 2026)
AI & society
-
Situational Awareness by Leopold Aschenbrenner (Blog, June 2024)
-
AI 2027 by Daniel Kokotajlo et al. (Forecast, April 2025)
-
The Art of Wanting by David Bau (Blog, January 2026)
-
Europe 2031 by Daan Juijn et al. (Forecast, June 2026)
-
AI 2040 by Thomas Larsen et al. (Forecast, July 2026)
-
Life on the Uncanny Precipice by Naomi Saphra (Blog, August 2026)
Software engineering
-
Why developers should also be product managers by Mark Saroufim
-
Does Computer Science Still Exists? by David Bau (Blog, March 2026)
-
Dealing with LLM Religion by David Bau (Blog, May 2026)
Academia and industry career advice
-
The Tao of PhD has a lot of relevant pointers specific to academia.
-
Advice for Faculty and Industry Interviews by Abhradeep Guha Thakurta (Blog, January 2025)
-
How to professor by Dirk Hovy (Blog, January 2025)
-
How I’ve run major projects by Ben Kuhn (LessWrong, March 2025)
-
How To Become A Mechanistic Interpretability Researcher by Neel Nanda (AI Alignment Forum, September 2025)
-
How to write gooder by Dirk Hovy (Blog, November 2025)
-
Conferencemaxxing: How to grow your profile and network as a scientist by Michael Saxon (Youtube, 2025)
-
How to write an okay research paper by Sasha Rush (Youtube, 2025)
-
Should I do a Postdoc? by Niloofar Mireshghallah (Youtube, 2025)
-
Behind the Scenes of my PhD by Benno Krojer (Blog, June 2026)
-
ML Job Interviews: The Ultimate Guide by Silvia Sapora (Blog, June 2026)
-
AI-native PhD Students by Eytan Adar (Blog, August 2026)
Comedy
- The Triestin Muleria: A Preliminary Characterization by Diego Manna (Bora.La, October 2009)2
-
The Company Man by Tomás B. (LessWrong, September 2025)
-
I, Sisyphus, am Ninety-five Percent of the Way There by Jack Loftus (McSweeneys, May 2026)
Courses
Interpretability
- Interpretability lectures by Atticus Geiger, Jack Merullo and Ekdeep Singh Lubana (Goodfire) at Stanford
- Neural Mechanics course by David Bau at Northeastern
- Mechanistic Interpretability course at Stanford
- LLM Interpretability course by Mor Geva et al. at Tel Aviv University
Other topics
- AI Research Experiences course by Pranav Rajpurkar at Harvard
- AI Safety course by Maksym Andriuschenko, Jonas Geiping et al. at University of Tübingen
- The Modern Software Developer course by Mihail Eric at Stanford
- AI Agents and Simulation course by Joon Sung Park (Simile) at Stanford
Books
You can check out my Goodreads favorites. I’m generally a big fan of sci-fi novels3 and non-fiction — mostly historical accounts, (auto)biographies and science-related books. As a young teen, my favorite books were the Zamonien collection by Walter Moers. These days, I mostly listen to books on Audible.
Music
I grew up mostly on Led Zeppelin and Linkin Park, and I even used to sing sometimes. These days, I listen to a bit of everything. Here’s some less mainstream pieces you are less likely to have heard before (especially if it’s not in your native language):
-
The Unsongs project by Moddi — English covers of songs that have been censored throughout history.4
-
alt-J — pretty much everything from An Awesome Wave and This is All Yours.
-
The Monolith of Phobos — Debut album by The Claypool Lennon Delirium.
Non-English
-
Rap – I’m a big fan of rap sonorities, and I mostly listen to artists that do not sing in English. Some of my favorites are Caparezza, Mike Shinoda, Watsky, TopGunn, Koriass and Loud.
-
Italian music — Fabrizio De André is a staple, and I enjoyed the commemorative album Faber Nostrum. La Niña is also great for Neapolitan evocative rhythms.
-
Québecois folk — Les Cowboys Fringants, especially La Grand-Messe, and Les Colocs, especially Dehors Novembre.
-
Latin American vibes — Chavela Vargas5 and Calle 13, especially Multiviral.
-
Norwegian rock and folk — Kaizers Orchestra is great, especially Ompa til du dør. I also like Moddi’s Norwegian songs, Til Ungdommen and Heiemo og Nykken.
-
Frisian — Nynke Laverman is great, especially her Sielesâlt fado debut album.
-
Russian — I was initiated to Russian music with Любэ6 and Владимир Высоцки. Some of my favorites are Трава у дома, Звезда по имени Солнце and Крошка.
-
Chinese – I know very little Chinese music, but I like 大魚 (Big Fish) by 周深, 馬頔 (especially 孤岛) and 川子.
Movies and series
I’m admittedly not big on movies, and my favorites are mostly mainstream sci-fi stuff (Matrix, Arrival, Interstellar, Dune, Ex Machina, I am Mother), all Tolkien adaptations, and fantasy / sci-fi thriller series (Dark, Black Mirror, Westworld, Severance). A recent favorite less-known outside of AI circles is Pantheon.
Podcasts and video channels
AI and science — I mostly listen to Dwarkesh and the 80,000 hours podcast, and watch videos from Kurtzgesagt (mostly cosmology), 3Blue1Brown (mostly math) and Rational Animations (mostly AI safety).
History — my favorite podcasts are both in Italian: Qui si fa l’Italia (modern Italian history) and Il podcast di Alessandro Barbero (mostly medieval/reinassance history). I also occasionally watch Nightshift.
Geopolitics — I really like the diverse selection of RealLifeLore and the China-focused analyses of Polymatter. I also like the curated documentary-like format of fern and hoog.
Quotes
Science and understanding
The first principle is that you must not fool yourself – and you are the easiest person to fool.
– Richard P. Feynman7
We look at the world once, in childhood. The rest is memory.
– Louise Glück
If you’re thinking without writing, you only think you’re thinking.
– Leslie Lampart
A process cannot be understood by stopping it. Understanding must move with the flow of the process, must join it and flow with it.
– First Law of Mentat, Frank Herbert’s Dune
Software engineering
Up until a few months ago, the best developers played the violin. Today, they play the orchestra.
– Justin Searls
Software 1.0 easily automates what you can specify. Software 2.0 easily automates what you can verify.
– Andrej Karpathy
We build computers and programs for many reasons. We build them to serve society and as tools for carrying out the economic tasks of society. But as basic scientists we build machines and programs as a way of discovering new phenomena and analyzing phenomena we already know about. Society often becomes confused about this, believing that computers and programs are to be constructed only for the economic use that can be made of them (or as intermediate items in a developmental sequence leading to such use). It needs to understand that the phenomena surrounding computers are deep and obscure, requiring much experimentation to assess their nature. It needs to understand that, as in any science, the gains that accrue from such experimentation and understanding pay off in the permanent acquisition of new techniques; and that it is these techniques that will create the instruments to help society in achieving its goals.
– Turing award lecture from Allen Newell and Herbert A. Simon (1975 Turing Award Winners)
Deep Learning
Reinforcement learning is sucking supervision bits through a straw.
Andrej Karpathy on the Dwarkesh Podcast
Footnotes
-
A response to A Pragmatic Vision for Interpretability to advocate for “ambitious interpretability”. ↩
-
This will make sense only if you are from Trieste. ↩
-
I’m especially fond of short stories. My favorite authors are Ken Liu (The Paper Menagerie and Other Stories) and Ted Chiang (Exhalation and Stories of Your Life and Others). ↩
-
I had the pleasure of seeing Moddi live in Groningen in 2021. ↩
-
Surprisingly, one of the best covers of her famous “Paloma Negra” is in Serbo-croatian. ↩
-
Especially true for interpretability researchers! ↩