Now running · Meeting 6 notes available

Étale Cohomology Seminar 2026

The seminar continues after the five-meeting pilot.

We follow Daniel Litt's Étale Cohomology and the Weil Conjectures. Videos, notes, references, and AI tools support individual preparation. We meet weekly to do what requires other mathematicians: question, verify, challenge, explain, review, and decide together what is worth remembering.

Fridays · 17:00-18:30 Lima · 19:00-20:30 Brasília Online via Google Meet Today, Friday, November 13, 2026: logistics meeting (no mathematical session)

An imaginative illustration of the conversations and ideas surrounding the development of étale cohomology.

Seminar notes

Meeting 6 notes are available.

Pushforwards, stalks, sheafification, and the abelian category of sheaves · Friday, November 6, 2026

Read Meeting 6 notes

What this is

A serious seminar, not another notes archive.

The seminar is a public, collaborative pilot for undergraduate, master's, PhD, and other mathematically mature participants who want to study the first part of Litt's course together.

The repository records what the community contributes beyond the lectures: examples, non-examples, verification, peer review, proof digestion, useful friction, attribution, and a small shared toolkit.

Read the full project README →

Why another seminar?

Explanations are abundant. Human mathematical judgment is scarce.

Videos + books + AI

explanation · information · examples · assistance

The seminar

verification · discussion · judgment · responsibility · shared understanding · mathematical community

We are not meeting because mathematical explanations are scarce. We are meeting because human mathematical attention, criticism, trust, and shared understanding are scarce and valuable. Our philosophy is inspired in part by questions raised in Terence Tao's Mathematics in the Age of AI.

Read our full seminar values →

How it works

3 + 2 + 1 + E + R + P

3

Three things you think you understood.

2

Two points of mathematical friction.

1

One question worth discussing with humans.

E

One example or non-example that clarifies a definition or theorem.

R

Review one mathematical contribution made by another participant.

P

Problem of the Week. Engage seriously with one common problem. A partial solution, failed approach, or useful observation may be as valuable as a complete solution.

candidate peer checked discussed shared toolkit

Candidate examples and proofs are proposed, checked by another person, discussed by the seminar, and only then may enter our shared toolkit.

Read the contribution guide →

Problem of the Week

Problems are for thinking, not merely for obtaining answers.

AI can often produce a candidate solution. Our goal is to understand why a solution works, where the key idea lies, what fails in unsuccessful approaches, and whether another mathematician can verify and explain it.

try get stuck use references/tools reconstruct peer review discuss digest

Examples matter

The best example is the one that makes the mathematics clearer.

Every participant is encouraged to commit the example or non-example they found most illuminating. It does not need to be the most sophisticated example in the room.

GitHub helps us preserve who contributed it, how it was checked, what sources were used, whether AI or software helped, and how the contribution changed after review.

See the repository →

Who can participate?

One mixed-level seminar.

You do not need to understand 100% of every lecture before coming to the meeting.

Undergraduate

Definitions, examples, intuition, and good questions.

Master's

Also proof architecture, hypotheses, and independent examples.

PhD and beyond

Also context, criticism, reviewing, connections, and mentoring without dominating.

Serious curiosity and preparation matter more than academic seniority.

Read the detailed expectations →

Timetable

The pilot followed Litt's first five lectures; the seminar continues every Friday.

  1. Meeting 1 Aug 28 · Weil conjectures and motivation. Materials in progress.
  2. Meeting 2 Sep 4 · Étale morphisms and introduction to sites. Read Meeting 2 notes.
  3. Problem session Sep 11 · Exercises from Litt's Lecture 2: standard étale morphisms, the power map on Gm, 2-out-of-3, and completed local rings. Review the exercises.
  4. Meeting 3 Sep 18 · Sites, sheaves on sites, examples of sites, and the étale and fppf topologies. Read Meeting 3 notes.
  5. Meeting 4 Sep 25 · Morphisms of sites and fppf descent I. Read Meeting 4 notes.
  6. Meeting 5 Oct 2 · fppf descent II and the category of sheaves. Read Meeting 5 notes.
  7. Continuation Oct 9 · Continuation of the previous lecture: fppf descent II and the category of sheaves. Meeting 5 notes.
  8. Descent examples Oct 16 · Discussion of examples of descent.
  9. Recess Oct 23 · No meeting (organizer giving a talk at USP).
  10. Discussion Oct 30 · Harald Helfgott on the recent OpenAI release, in particular the lack of citation.
  11. Meeting 6 Nov 6 · Pushforwards, stalks, sheafification, and the abelian category of sheaves. Read Meeting 6 notes.
  12. Logistics meeting Nov 13 · No mathematical session; the group meets to organize the seminar.
  13. Nov 20 To be announced.
  14. Recess Nov 27 · No meeting (organizer travelling in Mexico).
  15. Meeting 8 Dec 4 · Čech cohomology, the Čech-to-derived spectral sequence, quasi-coherent cohomology in the étale and fppf topologies, and the Artin–Schreier sequence. Read Meeting 8 notes.

Every Friday, 17:00-18:30 Lima / 19:00-20:30 Brasília. Online via Google Meet.

After the five-meeting pilot, the group decided to continue with Litt's course.

Each week

Prepare individually. Verify together.

Before

watch → read → try the problem → use tools → contribute → review

During the 90-minute meeting

0-15
collective reconstruction
15-30
friction board
30-50
common Problem of the Week
50-70
example / theorem / proof deep dive
70-80
what should survive?
80-90
connections, community, next week

The opening reconstruction is AI-free and notes-closed.

AI and tools

AI is welcome. Responsibility remains human.

We use AI and mathematical software freely for preparation, while keeping human responsibility, verification, reviewing, attribution, and collective digestion at the center of the seminar.

  • Disclose substantial tool use.
  • Make contributions easy for other humans to review.
  • Human responsibility remains with mathematical claims.
  • Properly attribute mathematical sources and other people's ideas.

"GPT told me so" is not mathematical verification.

Read the values →