This website was built
in 30 minutes.
You can do the same!

Sundai.Club

MIT IAP 2026 · January 20–27

How to Ship Almost Anything with AI v2

Class number6.S093
LevelUndergraduate
Units6
GradingP/F
DatesJan 20-27
Time10am-4pm
Seats40
Quick links:DescriptionScheduleGradingFAQCheck out the recordings of the previous year
Course Materials
Access all the course materials here.

Rolling updates—check back regularly for new content.

Description

How to become the one-person unicorn?

Sam Altman predicts that soon there will be a billion-dollar company run by a single person. With agentic AI tools improving every month, it's becoming easier to believe. At Sundai, we've been building with AI every Sunday for the past 100 weeks, and in this class, we'll teach all the technical skills we've learned to build full-stack products single-handedly.

In this class you will build a fully autonomous agent to manage your/your company social media based on your 2nd brain app of choice. Gain hands-on experience with the latest tools and techniques, and level up your ability to ship.

What you'll learn

After taking this course, you will:

  • Understand how to design and build autonomous agentic systems with tool calling
  • Be proficient in using and fine-tuning multimodal AI models (LLMs and image generation)
  • Know how to implement RAG systems and semantic search for AI memory
  • Be able to deploy AI applications to cloud infrastructure
  • Understand best practices for web scraping and social media integration
  • Confidently use AI-powered developer tools for rapid prototyping
Who should apply?

Prerequisites:

  1. Good knowledge of Python (or at least one other programming language)
  2. Be comfortable with GIT
  3. (nice to have) Experience with JavaScript
  4. (nice to have) Basic knowledge of Machine Learning (i.e. 6.390)
Course Schedule
Jan 20TuesdayDay 1: Agent Logic & Content Selection
  • Lecture 1: Introduction and motivation for autonomous AI agents
  • Lecture 2: AI models fundamentals — tokens, long context, multimodal capabilities, model selection, self-hosting vs cloud, fine-tuning considerations
  • Workshop: Introduction to vibe-coding; design agentic architecture with tool calling (Pydantic); build post generator with system prompts; integrate with social media API
  • Deliverable: Python script that makes a post on a social media site based on system prompt about the company
Jan 21WednesdayDay 2: Multimodal AI
  • Lecture 3: How to use and fine-tune image generation models
  • Lecture 4: Building agentic systems with tool use; introduction to Claude Code skills
  • Workshop: Advanced vibe-coding techniques; fine-tune a diffusion model on custom images; integrate image generation with the agent; (optional) add Telegram approval workflow
  • Deliverable: Python script that posts to social media with AI-generated images of the company's mascot + (optional) pre-approval functionality
Jan 22ThursdayDay 3: Backend & Scraping
  • Lecture 5: How to design and build a backend
  • Lecture 6: Deploying your app to the cloud; introduction to web scraping
  • Workshop: Set up project structure; initialize SQLite database; build social media scraper with Playwright; store scraped content with metadata; create basic CLI for testing
  • Deliverable: Deployed Python script running on a schedule + (optional) scraper for relevant posts and automated engagement scheduling
Jan 23FridayDay 4: 2nd Brain & Memory
  • Lecture 7: Benchmarking, LLM as a judge, observability
  • Lecture 8: RAG and memory systems
  • Workshop: Implement comment monitoring; build response generation agent; create RAG system for post history; add vector embeddings for semantic search; implement memory and context management
  • Deliverable: Connect agent to note-taking system (text files, Notion, or Obsidian) with RAG implementation
Jan 26Monday[Optional] Day 5: Frontend Development
  • Lecture 9: Fast prototyping of frontends; basics of HTML, CSS, and JavaScript
  • Lecture 10: Introduction to modern frontend frameworks — Tailwind, Next.js
  • Workshop: Build a dashboard for monitoring and controlling the AI agent
Jan 27TuesdayLive-demo Presentations

Final project demonstrations with live deployment showcases

Grading

Grading is P/F. Hit the targets below and you're golden:

20%
Attendance/participation
40%
Daily Deliverables
40%
Final Project
Daily Deliverables

Each course day includes a deliverable that builds progressively toward the final project. Students are expected to complete deliverables during class time with TA support. By the final day the students are expected to implement one additional feature on top of the agent. Students will demonstrate their working systems with live deployments during the final presentation session.

Final Project

Students will build an autonomous AI agent that manages social media for a company or individual. The system will:

  • Monitor a knowledge base (text files, Notion, or Obsidian) for updates
  • Generate contextually relevant social media posts autonomously
  • Create custom images using a fine-tuned diffusion model based on the company's mascot
  • Scrape and monitor social media for relevant conversations
  • Engage with online discussions related to the company's domain
  • (Optional) Implement human-in-the-loop approval via Telegram
  • (Optional) Build a web dashboard for monitoring and control
Frequently Asked Questions

Can I take the class?

MIT undergrads and grads are welcome, plus students from cross-registration partners like Harvard and Wellesley (check with your registrar). Space is capped at 40, so apply early.

What is the time commitment?

Core programming runs Jan 20–27, 2026 with daily sessions 10am–4pm plus optional evening build rooms. Tuesday the 27th is demo day. Plan to ship something meaningful—you'll get out what you put in.

Can I attend remotely?

We record everything and share materials, but in-person participation is required for MIT credit and for the best build experience.

When will the lecture lineup be posted?

We're finalizing speakers now. Expect the full agenda (with guests from teams like Google and Stripe) to drop in early January.

What should I bring to class?

Bring a laptop with a working dev environment, access to GitHub, and accounts on at least one LLM provider. We'll send prep instructions so you can hit the ground running.

Course Materials
Access all the course materials here.

Rolling updates—check back regularly for new content.