The library

Resources for every stage.

Start here, keep building, or go to the frontier. Every recommendation includes a plain-English summary before you decide to open it.

Or just come to a session

Find the right place to begin

Start with the essentials, browse the complete library, or go straight to the frontier material.

Start here

For understanding the idea, choosing one tool, and making something for the first time.

First-build guide What is vibe coding?Starter tools

The full library

Tools, free courses, deeper learning, agents, skills, articles, and other places to learn in Oslo.

Free courses Agents and skills Articles + LLM primer

Frontier

Notes, research, and guest lectures about the people and systems shaping modern AI.

Frontier Systems notes Papers behind the course Guest-lecture resources

Start here

What vibe coding actually means, from the people who named it.

Andrej Karpathy

The original “vibe coding” note

In early 2025, the AI researcher Andrej Karpathy fired off a short post describing a new way to write software: let the model do almost all the work and just go with the flow. The phrase he used, "vibe coding," caught on fast.

Simon Willison

Not all AI-assisted programming is vibe coding

Soon after the phrase took off, the developer and writer Simon Willison stepped in to clear up a growing confusion: using AI to help you code is not the same thing as vibe coding. He wanted to keep the two ideas separate.

MIT Technology Review

What is vibe coding, exactly?

As the phrase spread far beyond programmers, MIT Technology Review published a 2025 explainer to answer the obvious question for a general audience: what does "vibe coding" actually mean, and where did it come from.

The tools we use

You only need one to start. Pick whatever gets you building tonight.

Paid

Anthropic

Claude Code

Claude Code is an AI helper that lives in your terminal, the plain text window where developers type commands. You tell it what you want in normal language, and it works directly inside your real project files.

Paid

OpenAI

Codex

Codex is OpenAI's coding agent. It reads, edits, and runs your code, and you can reach it from your terminal, your editor, a desktop app, the web, and inside ChatGPT.

Free

Anysphere

Cursor

Cursor is a code editor with AI built into its core. It looks and feels like a familiar editor, but a model is woven through every part of how you write and change code.

Free

Lovable

Lovable lets you describe the app you want in plain language and get back a working web app that is already online. There is no setup to wrestle with first.

Free

GitHub

GitHub Copilot

GitHub Copilot is an AI assistant that works inside your code editor. As you type, it suggests the next lines, and you can ask it questions without leaving your work.

Free

Vercel

v0

v0 turns a written prompt into a front-end user interface. You describe a screen, and it produces React components you can copy straight into your own project.

Free ways to go from zero

No prior coding needed. These are the on-ramps we point beginners to.

Coursera

Vibe Coding Essentials

This is a beginner specialization on Coursera, made by Scrimba, that teaches you to build real apps by working alongside AI coding tools. You bring the curiosity. You do not need any coding background to start.

Datawhale (open source)

Easy-Vibe: your first modern coding course

Easy-Vibe is a free, open-source course from Datawhale that walks you from a first idea all the way to a shipped AI app. It is open to everyone, and the whole curriculum is out in the open for anyone to read and improve.

Open source

Vibe Code Source

Vibe Code Source is a free, open-source library of copy-ready prompts paired with practical guides. Think of it as a shared toolbox you reach into when you are building with AI and want a proven starting point.

When you want to go deeper

For when the vibes meet a real codebase and you want to understand what is happening.

Anthropic

Building effective agents

Anthropic published this guide in 2024 to answer a question many builders were quietly struggling with: when you put a language model in charge of doing real work, what is the simplest setup that actually holds up? Its core advice is calm and a little surprising. Most of the time, you do not need a fancy autonomous agent at all.

Andrej Karpathy

From vibe coding to agentic engineering

In this conversation with Sequoia and Stephanie Zhan, Andrej Karpathy describes how building software with AI is growing up. The playful, improvisational style he once called vibe coding is giving way to something more deliberate, where a person directs capable agents and stays responsible for the result.

arXiv

Vibe coding, the academic version

In 2025, researchers Advait Sarkar and Ian Drosos published the first empirical study of vibe coding, the new habit of building software mostly by talking to an AI rather than writing code by hand. Instead of guessing how people do it, they watched them do it and listened to them think out loud.

Build your own agents and skills

Two hands-on courses by Akshat: one on AI agents, one on the skills that make them better.

Akshat Jain

Skill Factory

A skill is a small folder with a SKILL.md file that teaches an AI coding agent a habit once, so it applies it every time. This free course teaches you to write ones that actually get used.

Akshat Jain

Agent Factory

An agent is an AI that reasons, acts, and observes in a loop until it reaches a goal, not just a chatbot that answers one question. This free 23-module course teaches you to build real ones.

Go to the frontier

Curated from Stanford's CS 153. The deep end, the papers that built modern AI.

Stanford

CS 153: Frontier Systems

A Stanford course that walks you up the whole ladder of modern AI, from the raw silicon in a chip to the policy debates happening in public, with people who actually built these systems stopping by to talk.

Krizhevsky, Sutskever & Hinton

AlexNet: ImageNet classification with deep CNNs

In 2012, three researchers trained a deep neural network on graphics cards to recognize objects in photos, and it won a major contest by such a margin that it changed how the whole field thought about machine learning.

Mikolov et al., Google

Word2Vec: Efficient Estimation of Word Representations

In 2013, a Google team found a fast way to turn words into lists of numbers that capture meaning, so that simple arithmetic on those numbers could answer analogies like king minus man plus woman lands near queen.

DeepMind

Playing Atari with Deep Reinforcement Learning

In 2013, DeepMind built a single program that learned to play many different Atari video games well, looking only at the raw screen pixels and the score, with no game-specific instructions.

Vaswani et al., Google

Attention Is All You Need

In 2017, a Google team introduced the Transformer, a model that drops the older habit of reading text word by word in order and instead lets every word look directly at every other word at once. It became the foundation of modern large language models.

Google

BERT: Pre-training of Deep Bidirectional Transformers

In 2018, Google released BERT, a language model that reads a sentence from both directions at once and learns from huge amounts of plain text first, so it can then be adapted to many specific tasks with little extra training.

OpenAI

Scaling Laws for Neural Language Models

In 2020, OpenAI researchers found that the quality of a language model improves in smooth, predictable curves as you give it more compute, more data, and more parameters, so you can forecast how good a model will be before you build it.

OpenAI

Language Models are Few-Shot Learners (GPT-3)

In 2020, OpenAI showed that a very large language model with 175 billion parameters could perform new tasks just from a few examples written into the prompt, without any retraining.

Ho, Jain & Abbeel

Denoising Diffusion Probabilistic Models

In 2020, three researchers showed a clean way to generate images by teaching a model to reverse noise: start from pure static and remove a little noise at a time until a real-looking picture appears.

OpenAI

Training Language Models to Follow Instructions (InstructGPT)

In 2022, OpenAI showed how to make a language model genuinely follow instructions by tuning it on human preferences, a method called RLHF that became the basis for ChatGPT.

DeepMind

Training Compute-Optimal LLMs (Chinchilla)

In 2022, DeepMind found that many large models had been built too big for the amount of data they were trained on, and that a smaller model fed more data can do better for the same training budget.

From the guest lectures

The work behind the people building the frontier: generative images, video, and voice.

Rombach, Blattmann et al.

High-Resolution Image Synthesis with Latent Diffusion Models

This 2022 paper is the one that put text-to-image generation in everyone's hands. It is the research behind Stable Diffusion, and its big idea was to make image generation cheap enough to run on a single ordinary graphics card.

Blattmann et al., Stability AI

Stable Video Diffusion

Released in 2023, this work takes the same recipe that made Stable Diffusion good at single images and stretches it across time, so the model produces short video clips instead of standalone frames.

Amit Jain

Luma AI

Luma AI is a company working on multimodal AI for video and 3D, led by Amit Jain. Its best-known product, Dream Machine, turns a prompt or an image into video, and the longer ambition is software that understands and simulates the physical world.

Mati Staniszewski

ElevenLabs

ElevenLabs is an AI audio company, co-founded by Mati Staniszewski, built around one focused mission: making synthetic speech sound genuinely human. Its tools cover natural text-to-speech, voice cloning, and real-time translation.

Sequoia · Training Data

Why voice will be the core interface

This is a 2025 podcast episode from Sequoia's Training Data series, featuring ElevenLabs CEO Mati Staniszewski. The conversation makes the case that voice is becoming a primary way we interact with computers, and that staying focused is how a company wins.

AI for everyone

Why we run this for free and in the open. Cheap, safe access should not be a privilege.

Anthropic

Core Views on AI Safety

This is the 2023 piece where Anthropic lays out, in plain terms, why the company was founded and what keeps its researchers up at night. It is the closest thing to a mission statement for how to build powerful AI without getting burned by it.

Rich Sutton

The Bitter Lesson

In this short 2019 essay, the reinforcement learning researcher Rich Sutton points to a pattern that keeps repeating across decades of AI. General methods that ride on raw computing power tend to win, and clever human-crafted shortcuts tend to lose.

Other places to learn in Oslo

We are not the only room in town, and that is a good thing.

Meetup

Oslo AI

Oslo AI is a long-running meetup where people in and around the city gather to talk about artificial intelligence in plain, practical terms. It is run by volunteers, it costs nothing to join, and it is one of the bigger AI communities in Norway.

AI Tinkerers

AI Tinkerers Oslo

AI Tinkerers Oslo is the local branch of a worldwide network for people who like to build things with AI rather than just talk about them. Expect demo nights, hackathons, and a hands-on, roll-up-your-sleeves mood.

Meetup

Vibe Coding Collective

The Vibe Coding Collective is an international community built around relaxed, social coding sessions where people build with AI in good company. It is friendly to beginners and leans more toward easy company than intimidating expertise.

New to this? Come build with us.

Reading is good. Building with people is better. Our drop-ins are free and open to total beginners.

RSVP for the next session