AI use policy

How AI helped make this workshop, and how you can use it during the day

Where AI tools are welcome in the tensors workshop, the three things that stay yours alone, and how AI was used to build the materials.

Español

In short

AI assistants are welcome during the workshop, with three exceptions: your first prediction, the Kahoot quizzes, and the entry and exit checks are yours alone. Whatever an assistant tells you, run the code and look at the shape: the notebook’s own check decides, not the assistant.

Using AI during the workshop

You may use any AI assistant, including Gemini inside Colab, to explain an error message, a NumPy function or a term, and to review code you have already written. Try each exercise yourself first, the same way you try it before opening its folded solution, and then ask for help.

Three things are yours alone

  • Your first prediction. Sections open on a predict-first question, and some on a Myth or fact? claim. Make your own guess before you run anything or ask anyone. A wrong guess you made yourself teaches you more than a right one you were given.
  • The Kahoot quizzes. Play them on your own, from memory.
  • The entry and exit checks. They measure what you understand, so the instructors can see where the room is. An assistant’s answer tells them nothing about that.

Check what it tells you

An assistant can sound sure and still be wrong, and it is wrong most often about what this workshop is about: which axis is which, and what shape comes out. Before you trust an answer:

  • run it, and print .shape;
  • compare it with the notebook’s expected result and with the handbook;
  • if the two disagree, trust the notebook’s check and ask in Discord.

The worked mistakes are good practice: each one is code that runs, prints something plausible, and is wrong.

Say when it helped

When your group posts its three-line share-back in Discord, or you share work from the workshop anywhere else, say whether an AI assistant helped, and how. This page does the same for the materials.

Keep private data out

Code and output from the workshop’s notebooks are fine to paste into an assistant. Do not paste anything that is not yours to share: data from your job, other people’s personal information, passwords or API keys.

How AI was used to make this workshop

  • Code and site. The notebooks, the interactive widgets, this site and the scripts that build and check them were written with AI coding assistants (Claude, from Anthropic), working under the instructors’ direction. Most commits in the repository name an AI co-author.
  • The Spanish handbook is a machine translation of the English one, which remains the source of truth.
  • The companion’s summaries, audio, videos and self-checks were generated with NotebookLM and have not been checked by a person.
  • What is checked automatically. Before a change reaches the site, an automated run executes the notebooks’ code and checks the numbers it prints. Nothing checks prose that way, in either language.
  • Who is responsible. The instructors chose what is taught and in what order, and they are responsible for every error, whoever or whatever wrote it first. If you find one, say so in Discord or write to ravkalia@gmail.com.