AI 4 Coding

Day 1: Course Overview

Eduard Bukin
ebukin@worldbank.org

Distributional Impact of Policies. Fiscal Policy and Growth Department

2024-04-20

Motivation

Chat-Based AI tools Claude | ChatGPT | Gemini | MAI

  • use familiar technologies: 🌐 Web-Browser, 💬 Chat, 📝 Stata editor, 📋 copy-paste
  • improve our workflow.

Is this the most effective way to use AI in coding❓

There are many Integrated Development Environments (IDEs) for data science with AI integrated into your coding workflow. Introducing Positron with Github Copilot is the goal of this course.

See also: previous seminar on AI-Assisted Coding in Stata and R on GeoPovlearn.

Course Overview

Goal: Equip economists with AI-skills for coding.

2 × 3-hour sessions Self-study in between GitHub Copilot · Positron IDE Stata & modern stat languages

Understand

  • Core LLM concepts: tokens, prompts, context windows, agents
  • How these affect AI-assisted coding
  • Strengths and limitations for analytical workflows

Practice

  • Use AI to understand, revise and reuse code
  • Plan execution and write prompts for multi-step tasks
  • Leverage #tools, skills and /agents
  • Supervise and course-correct AI-generated code

Take Control

  • Data-security practices and guardrailing AI.
  • Responsible-use guidelines for AI-assisted

Course Structure

Day 1

Basics

  • Software setup & prerequisites
  • AI concepts: tokens, context, agents
  • Demo of AI-assisted coding
  • Common workflows & best practices

Self-Study

The most important part

  • Build the habit with your own code
  • Structured case studies to guide you
  • Come back with questions & observations

Day 2

Go deeper, work smarter

  • Debrief self-study experiences
  • Planning & executing complex tasks
  • #Tools, skills, /agents, /prompts
  • Context engineering & frontier techniques

Logistics

Morning Evening (repeat)
Format In person + virtual Virtual only
Session 1 Mon Apr 20, 9 AM–12 PM ET Mon Apr 20, 9 PM–12 AM ET
Session 2 (repeat) Thu Apr 23, 9 AM–12 PM ET Thu Apr 23, 9 PM–12 AM ET
  • Questions during session: use Teams chat or raise hand please

Agenda

Part 1 — Setup & Live Demo (80 min)

Time Topic
10 min Welcome, introductions, course overview
30 min Software overview: Positron, GitHub Copilot, and Stata setup
40 min AI in action with Stata (R): my typical data workflow using AI

☕ Break (10 min)

Part 2 — Concepts & Practice (90 min)

Time Topic
30 min AI overview presentation: how GitHub Copilot and LLMs work, key concepts, and capabilities
30 min Cookbook: securing sensitive data, guardrails, and responsible AI use
30 min Self-study exercises overview and Q&A

Resources