Intro To Agentic Programming
UniTo Master in Economics
1 Welcome to Introduction to Agentic Programming
- This website is the home of the course Introduction to Agentic Programming.
- This course assumes no prior programming experience.
- Below you will find setup instructions and a syllabus.
- You can obtain all material for this course from the associated github repository at link
1.1 Objectives of this Course
- This course is about programming in the age of coding agents. After this course, you should be able to supervise an AI agent to do real, reproducible research work — not just write code by hand.
- You will learn the layered model of modern AI tooling: model (the LLM), harness (the tool driving it, e.g. OpenCode), tool calls, and skills — and why “programming” now means verifying an agent’s output, not only producing it yourself.
- You will build the habit, repeated across every session: where could this be wrong, and how do I check?
- We work in both Python and R. No prior programming experience is assumed.
- The harness used in class is OpenCode with a free-tier model provider, so nobody is blocked by needing a paid subscription.
2 Program
The course is 12 in-person/remote sessions (S1–S12) plus an async setup session (S0). See Section 3 for S0.
| Session | Mode | Topic | File |
|---|---|---|---|
| S0 | async | Accounts & install | Prework |
| S1 | in-person | Kickoff: model/harness/tool/skill, shell, first git repo | Kickoff |
| S2 | in-person | Programming constructs, agent-native (R & Python) | Constructs |
| S3 | remote | Git + project config as agent-compatibility infrastructure | Git & Project Config |
| S4 | in-person | How LLMs actually work | LLM Internals |
| S5 | in-person | How agents work: context, tools, harnesses | Agents: Context, Tools, Harnesses |
| S6 | in-person | Data wrangling with DuckDB | DuckDB |
| S7 | remote | Testing & verification I: unit tests | Unit Tests |
| S8 | in-person | Testing & verification II: verification patterns | Verification Patterns |
| S9 | in-person | Cost & model economics | Cost & Model Economics |
| S10 | in-person | Spatial data, agent-native | Spatial Data |
| S11 | remote | Reproducibility & failure modes | Reproducibility |
| S12 | in-person | Capstone project | Capstone |
I just want to run some python code right away!! 👉 https://github.com/floswald/hello-devcontainer-with-jupyter
2.1 Schedule
We meet every Tuesday, 11:00–14:00, Laboratorio Informatico C. Some meetings cover two sessions back-to-back (longer), others one (shorter) — see each session’s linked file for content.
S0 (accounts & install) was meant as async pre-work, but since it slipped, we’re doing it together today, Tue 6 Oct 2026, before S1 starts next week.
| Date | Sessions | Time (used) | Room |
|---|---|---|---|
| 6 Oct 2026 | S0 | 11:00–12:30 | Laboratorio Informatico C |
| 13 Oct 2026 | S1 + S2 | 11:00–14:00 | Laboratorio Informatico C |
| 20 Oct 2026 | S3 | 11:00–12:30 | Laboratorio Informatico C |
| 27 Oct 2026 | S4 + S5 | 11:00–14:00 | Laboratorio Informatico C |
| 3 Nov 2026 | S6 | 11:00–12:30 | Laboratorio Informatico C |
| 10 Nov 2026 | S7 + S8 | 11:00–14:00 | Laboratorio Informatico C |
| 17 Nov 2026 | S9 | 11:00–12:30 | Laboratorio Informatico C |
| 24 Nov 2026 | S10 + S11 | 11:00–14:00 | Laboratorio Informatico C |
| 1 Dec 2026 | S12 | 11:00–12:30 | Laboratorio Informatico C |
3 Setup Instructions
- You must bring your own laptop to each class.
- Please make sure you have an up to date operating system, i.e. run a software update before we start.
- I put together a suite of extremely detailed installation instructions. You should follow those here before coming to the first session in class!
- Each Operating System has different instructions to follow.
4 Code of Conduct
If you decide to participate in this course, I expect you to abide by the following minimal code of conduct.
- Be polite to the other class participants.
- While in class, do not spend time on messaging apps, chat rooms, computer games, or similar content.
You can expect your instructor to abide by the same code of conduct, so this is a matter of mutual respect. If you are found in breach of the above you will be given a single warning, and I will ask you to no longer join the course after a second time. Your grade will be “fail”.
5 License
All lectures of this course are derived from the work of the Software Carpentry. Their material is licensed under creative commons license 4.0, whereby I am allowd to share and remix the content, if appropriate attribution is given.
Those terms apply to anyone wanting use material on this website as well. Thank you.
© Florian Oswald, 2026