Gentle Intro to Modern, Agentic Workspace

From Manual Labor to Autonomous Ecosystems

Abstract

In the past several months, the way we interact with technology has fundamentally shifted from simple chat prompts to the creation of autonomous ecosystems. This isn't just hype—we are seeing tasks that previously required an entire engineering team and a month of development being completed in a single hour on a personal laptop.

This mini-course offers a low-barrier entry into this "agentic" shift. We move beyond basic AI interactions to show how research and teaching teams can tame legacy codebases, migrate pipelines safely, and stand up classroom-ready workflows like synchronized assignments and supervised browser grading. The goal is to transform your digital environment from a static folder of files into a dynamic, agent-driven workspace that collaborates with you across lab and lecture contexts.

Repository Contents

File	Description
slides.md	Marp presentation source file
project-description.md	Project goals and workflow description
README.md	This file

Core Concepts

The 4 Quadrants Framework

	Agent Knows	Agent Doesn't Know
You Know	Reviewer (accelerate)	Teacher (provide context)
You Don't Know	Student (learn)	Explorer (discover)

Workshop Scenarios

This workshop explores the following real-world scenarios:

1. Legacy Code Organization — Taming messy legacy code and metadata
2. Library Migration — Migrating from dependency-heavy code to vanilla implementations
3. Assignment Consistency Automation — Keeping assignments in sync across lectures, LMS, and answer keys
4. Autonomous Browser Grading — Supervising AI to grade assignments in the LMS with an audit trail

Best Practices

1. AI as Echo-Chamber — AI is an echo chamber of infinite knowledge and experience; ask boldly
2. AI-Native Approach — Let AI suggest solutions and find the right tools
3. Turn One-Time to Rules — Convert one-off requests into reusable rules, workflows, or skills
4. Manage Long Context — Use planning documents, checkpoints, and voice input to manage long contexts
5. Constantly Verify — Continuously check understanding by requesting summaries; ensure AI stays on the same page
6. Multi-Agent Cross-Check — Use multiple agents to cross-check quality, but allow only one to modify code

Workflow

Markdown → Marp → PDF/HTML

Presentations are built with Marp and can be exported to PDF or HTML.

Getting Started

1. Clone this repository
2. Install Marp CLI or use the Marp extension for VS Code
3. Preview or export slides.md

# Export to PDF
marp slides.md -o slides.pdf

# Export to HTML
marp slides.md -o slides.html

License

MIT