When you first arrive at university, you may not yet know what you want from the next few years. You hear about graduate admissions, competitions, and joining labs, but may still wonder how these relate to your future—and why you would want to do research.
Or perhaps you already want to begin. We asked these questions too: how has everyone already joined a lab when I have never learned C? I have watched courses—what can I actually do next? I want to contact a professor, but what happens after I send the message?
This guide is for students starting university who want to explore AI and research. It brings together our answers to questions about university choices and getting started in research, alongside beginner projects, key papers across research areas, courses, and tools. It also covers contacting faculty, joining a lab for the first time, failed experiments, and the anxiety of comparing yourself with your peers.
Before you begin
A letter to new university students
What choices does university offer, and where do you want to go? I begin with my own uncertainty as a new student, then talk about grades, research, the people I met, and life outside research.
How to get started in research
What does research involve, and why try it? From choosing a direction and contacting faculty to reading papers, running experiments, and submitting your work, with questions about gaps in your knowledge, falling behind, and failed experiments.
AI, a history
From Dartmouth to language models, generation, robotics, and scientific discovery: the ideas that changed AI, and how their paths meet.
Getting started with AI
Your first AI project: where to begin?
Train a model to recognize handwritten digits, or detect people and cars in your own photos. Start with MNIST or YOLO and work through running, training, and modifying the code. There are also beginner projects for agents and robotics.
What do different AI research areas study?
Language models, agents, vision, multimodal learning, generation, RL, world models, embodied AI, and systems. Each starts with a question, then connects papers, explanations, and projects.
Which foundations do you need, and how much?
Python, mathematics, ML/DL, computing, and architectures. Learn the concept blocking you, then return to see how it works in your project.
Questions that come up during research
Finding and reading papers · Understanding experiments · Writing, figures, and talks
Contact and collaboration · Working with AI · Publication
How do you read your first paper, interpret experiment results, or write your first message to a professor? Methods and tools are organized around the research process, so you can return when a specific question comes up.
Experience, life, and perspectives
Experience, practical lessons, and wrong turns
From my first year to a first rejection, alongside other people's experiences and judgments.
Living well
Grades, research, friends, and hobbies all belong in university life. These notes begin with personal experience.
Lookout: what are researchers and labs working on?
Start with work that makes you curious. Meet the people and labs behind it, then follow papers, code, and discussions.
Research communities and knowledge sharing
OpenEnvision
OpenEnvision (OE) is an open AI research community connecting academia and industry, with interests in world models, multimodal intelligence, vision, and embodied AI. It also shares research knowledge through curated writing, interviews, and courses.
- BlogrXiv: AI research blogs and technical writing brings together research blogs, lab essays, and technical notes. Browse by field for explanations, engineering experience, and research methods, then follow links to the original articles.
- ScholarTube: AI interviews, podcasts, and courses collects long-form researcher interviews, video podcasts, complete courses, and research talks across agents, world models, vision, robotics, and research practice, with links to the original videos.
You can also recommend articles and videos to BlogrXiv and ScholarTube.
Lumina
Lumina brings together embodied AI research, open projects and community events. Its Embodied AI Guide organizes the field’s learning resources; Talks, research coverage and events on the website introduce the people and projects behind the work.
AgentHub
AgentHub collects discussions from the Agent community into daily and weekly digests, covering agent research, tools, engineering practice and industry developments. Browse by date, discussion group or external news to see what people are working through, how they compare technical approaches and where their views differ.
Resource index
The resource index collects courses, paper lists, author blogs, LessWrong, research teams, and Q&A. If you already have a question, search for Mu Li, RoPE, Scaling Ladder, or contacting a supervisor.