Start here: this week
Everything that is graded, in one place
Two kinds of grading, and they never mix:
| The track | The final assignment | |
|---|---|---|
| What | 15 marked items: sessions 1 to 14, and the cap01 notebook |
your research agent, on 15 private questions |
| Worth | 5,100 marks in all | a pass or not, and the certificate |
| Where you see it | the track, one row per student | dev3pack-submissions/finals/<you>/result.json |
| How you hand it in | uv run bootcamp submit <item> --github <you> --push, from your course folder |
uv run bootcamp final submit --github <you>, from my-final-assignment |
The Gecko capstone, your store and buyer, is neither: it is judged from your own
my-gecko-buyer, and you hand in a link to it by 5 October. Session 15 has no
notebook. Both deadlines, step by step: hand in by 5 October.
The 15 marked items
Every item is handed in the same way, from your course folder, after git pull and
uv sync --extra projects --extra agents:
uv run bootcamp check ch07 # see what passes, before you hand in
uv run bootcamp submit ch07 --github <you> --push # hand it in
Replace ch07 with the item. Save the notebook first: submit reads the file as
saved on disk. You can hand an item in again at any time.
Week 1
| Item | Session | Exercises | Marks | Open |
|---|---|---|---|---|
ch01 |
1. Configure the assistant | ch01-e1 |
100 | page · notebook |
ch02 |
2. Call a model through the adapter | ch02-e1 to e4 |
400 | page · notebook |
ch03 |
3. Structured outputs | ch03-e1 to e3 |
300 | page · notebook |
ch04 |
4. Bounded tools | ch04-e1 to e3 |
300 | page · notebook |
ch05 |
5. A deterministic mini-agent, plus the week 1 challenge | ch05-e1, ch05-e2, and up to 500 |
700 | page · notebook · challenge |
Week 2
| Item | Session | Exercises | Marks | Open |
|---|---|---|---|---|
ch06 |
6. A retrieval baseline | ch06-e1 to e3 |
300 | page · notebook |
ch07 |
7. Retrieval and grounding metrics | ch07-e1 to e3 |
300 | page · notebook |
ch08 |
8. Loops and graphs | ch08-e1 to e3 |
300 | page · notebook |
ch09 |
9. Trace and evaluate an agent | ch09-e1, ch09-e2 |
200 | page · notebook |
ch10 |
10. Skills and an ADR, plus the week 2 challenge | ch10-e1, ch10-e2, and up to 500 |
700 | page · notebook · challenge |
cap01 |
The final assignment's notebook | cap01-e1 to e5 |
500 | page · notebook |
Week 3
| Item | Session | Exercises | Marks | Open |
|---|---|---|---|---|
ch11 |
11. State and memory | ch11-e1 to e3 |
300 | page · notebook |
ch12 |
12. MCP architecture and primitives | ch12-e1 to e3 |
300 | page · notebook |
ch13 |
13. Build and secure an MCP server | ch13-e1, ch13-e2, ch13-e4 |
300 | page · notebook |
ch14 |
14. Deploy and operate | ch14-e3 |
100 | page · notebook |
The exercise ids are the ones the checks use. Session 13 has no e3, and session 14's
one exercise is e3: that is how they are numbered, not a missing exercise.
The three that need more than the notebook
The weekly challenges (ch05 and ch10, up to 500 each). You build the challenge in
its demo notebook: demo 8 for week 1,
demo 10 for week 2. Save it after the line
week 1 challenge: N/500 (or week 2 ...) has printed. Then
uv run bootcamp submit ch05 --github <you> --push carries the saved demo with your
hand-in, and the track reads your points from that line. The demo as shipped earns
nothing: the points need code that is yours.
cap01, the final assignment's notebook (500). It lives in the course folder, under
unit 2. It is scored on the track like a session. It is not the final assignment's
grade: that is the 15 private questions below.
The final assignment tutorial shows how cap01 connects to
your repository.
ch01, session 1 (100). It changes settings on your own machine, so the course
cannot re-run it for you. Hand in the notebook as it says; the track counts the
hand-in.
Reading the track
Your row on the track
shows each item as earned/possible. An empty cell means nothing was handed in yet,
which is not the same as a zero. 200/700 on ch05 or ch10 means the session is
done and the challenge is not. 5,100/5,100 is every item, including both challenges.
A pull request that is green merges on its own, and the track updates after the merge. No checks, or it never merges? Handing work in has what to do.
The final assignment: graded separately
Yes, it is graded, and it is the one that earns the certificate. It is not a column on the track.
- What: your research agent answers 15 private questions, 6 of them critical.
- Pass: a score of at least 30% and every critical question passed.
- Hand in: from
my-final-assignment,uv run bootcamp final submit --github <you>. Rehearse first with--dry-run. - Your result:
dev3pack-submissions/finals/<you>/result.json, a few minutes after the pull request merges. The latest submission counts.
Every step: the final assignment tutorial.
Share it
Your work is public on purpose, so it is worth showing:
- Your track row. The track is public: link it and point at your row. Each score
sits next to the notebook that earned it, in
submissions/<you>/. - Your final assignment repository. It is your showcase: its README shows a demo, your measured results and the command that produced them (section 12 of the tutorial).
- Your certificate, once the course issues it. It is signed, and anyone can verify it: the final assignment page shows how. Add it to your LinkedIn profile under Licenses and certifications, and post it with the link to your repository.
- Your capstone presentation. A two-minute clip of your buyer refusing the judge's card, with the explorer link to your devnet receipt, says more than any badge.