method where you build apps, websites, or games by describing what you want in plain English (or natural language) to an AI tool, rather than writing code manually. 3
Includes as a sidebar in your IDE: • Write prompts • Limited Access: • Low access to source code. • Only know about the files loaded in editor. • User discretion on accepting AI generated code. • Copy and paste code into IDE. • Need to know Programming Syntax 5
Generation • AI can do research & look at documents to gain context and knowledge. • Tool Calls • Call 3rd Party interfaces (API or CLI) to help with research • Model Context Protocol • A way to do Tool Calls. • Agent to Agent Protocol • Agents talking to one another. 6
Write Good Prompts 7 Watch Context Window • Write good prompts. • • • Use clear English. Sometimes it’s a conversation to refine your ideas. Watch your Context Window. • Use /compact or start new session if filled up.
coding agent. It’s available as a terminal-based interface, desktop app, or IDE extension. • The primary way we will be using OpenCode today is the Terminal User Interface (TUI). 12
Generate and edit your application using AI-powered VIBER Studio — from requirements to a running app. • Vibe-code your project with a full IDE, AI agent, and live preview. • Prerequisite: • TechPass (https://portal.techpass.gov.sg) • Platform AI (to get an API Key) (https://platform.ai.tech.gov.sg) http://go.gov.sg/stack2026-vibe-coding-vi ber (Open this link in your COMET laptop browser) 13
JY Lim • Rudy Kurniawan • Gina Chiang • Sun Huiyuan • Heng Li Meng • Tan Siong Chai • Kim R. Casper • William Afendy • Lai Johnson • Wu Huijie • Lo Jun Ting Kenneth • Yeo Kim Hock • Megan Lee • Yeo Su Jie
• 5 Coding Exercises http://go.gov.sg/stack2026-vibe-coding-viber (Open this link in your COMET laptop browser) https://go.gov.sg/stack2026-vibe-coding-workshop-materials 15
ARCHITECTURE 02. PROCESSING UNITS What is a Large Language Model (LLM)? What is a Token in LLM? • Use appropriate models for tasks. • Use more powerful thinking models for tasks that require better planning. • Use cheaper models for mundane tasks. • Tokenization is the process of breaking down text into smaller, manageable units called tokens. • These tokens can be words, parts of words (subwords), or even individual characters. 03. MEMORY & CAPACITY 04. EXTENSIONS & CAPABILITIES What is a Context Window? What are skills? • To prevent memory loss, always remember to compact. 17 • Things that you keep repeating, wrap them up into a Skill. • Reusable across sessions. • Shareable with others - just a bunch of Markdown files.