AI Tutorials for Companies & Teams

A little cheat sheet for you and for us: just the tools and info you actually need right now (changing rapidly and almost daily, don't get attached to any tools)

Hands-on Setup:

Many coders love working directly in the command line. We prefer VS Code for almost everything, because you get code, files, terminal, and AI plugins cleanly in one place.

Google Antigravity

Agent-first IDE from Google. Runs as a native development environment on your machine, not in the browser.

GitHub Copilot

Currently the best price-performance ratio for many. Students often get it free, and you can compare multiple models directly in the editor.

Claude

Strong in chat for structure, explanations, and longer coding tasks.

Qwen Code

Best via the VS Code plugin if you want to try Qwen directly in the editor.

VS Code

Our recommended setup. Codex, GitHub Copilot, Claude, and Qwen run here in one place with files, Git, and terminal.

Codex

If OpenAI, use Codex instead of ChatGPT for coding. Strong for agentic work directly in the tool and editor.

✍️ Prompt Tips for Beginners:

Describe the goal, user, platform, and desired result as concretely as possible.
Currently, it's best to prompt and code in English. German also works but usually a bit worse. Most models are primarily trained on English, and the programming world's language is English.

🧱 Especially Useful for Beginners:

Start with a small project rather than a huge idea.
First, let the model build a simple file and folder structure.
Ask the model to briefly explain each step and only make small changes at a time.

✅ Good First Projects:

Small landing page
To-do app
Notes app
Simple data or CSV viewer

🧪 Good Working Method:

First let it plan, then implement step by step.
Briefly test after each step.
Copy errors directly back into the tool and ask for a concrete fix.

🧠 Integrating Local LLMs:

llama.cpp, Ollama, or LM Studio let you bring local LLMs into VS Code or your own setups.

📁 What You Almost Always Need:

Git or GitHub for version control
A terminal for commands
A good editor like VS Code
A clear README with goal and setup

📱 Mobile Cross-Platform Apps (iOS + Android):

React Native Framework
Expo Runtime

🖥️ Local Cross-Platform Apps (macOS + Windows):

Tauri Framework

🍎 Local macOS Apps:

Native Swift Framework
Xcode Project

🧠 Local LLMs:

Everything runs locally on your machine. Slower than the AI providers, but very secure and costs only electricity, no subscriptions. Almost all local LLM programs are based on llama.cpp, an important open-source library.

Ollama

Very simple llama.cpp wrapper. A bit like ChatGPT-style for local models.

LM Studio

Slightly more complex, but still simple wrapper with more configuration options.

llama.cpp

For pros who want to integrate local LLMs into their own tools or work directly with the command line.

Hugging Face

Database for current and distilled models.

🧪 Distilled Models:

Smaller, trimmed versions of large models. They're faster, cheaper, and often good enough for many practical tasks.

📦 Typical Sizes:

Simple: 3B to 7B, often ~2 to 5 GB
Medium: 8B to 14B, often ~5 to 10 GB
Advanced: 32B and up, often 20 GB+

Rule of thumb:
16 GB RAM or VRAM is the absolute minimum.
More memory usually means larger models, more context, and smoother work.

⚠️ Requirements:

Computer with at least 16 GB RAM
or a strong Nvidia GPU
The more memory, the better.

🧩 Key Terms:

🔢 4B, 6B, 8B etc. roughly indicate the number of parameters in a model.
More parameters usually means more knowledge, nuance, and complexity. But it also requires more memory and compute.
🪟 GGUF is the standard format for many local llama.cpp setups and usually the best choice on Windows.
🍎 MLX is especially interesting for Apple Silicon and local models on Mac.

🏆 Best Local Coding Model: Qwen3 Coder Next (45 to 85 GB).

Other models we use locally for coding: Qwen, Ministral, and GLM.

🎥 For image and video processing we use:

Qwen and SmolVLM.

🚀 Your first own Vibe Coding project:

1. Open VS Code, click Extensions on the left and install Claude and ChatGPT.
2. Log in with your account. It has to be a paid account, otherwise you cannot code with it.
3. After logging in you see the chat window on the right. You are ready to go.
4. Create an empty folder and open it in VS Code. That is your project.
5. First tell the chat to create an md file (a simple text file) for your idea, and hand it the ground rules below.

Then copy one of the three project prompts underneath. Everything you need to copy is right here on this page.

1. KISS (Keep It Simple, Stupid) – keep it simple:

The simplest solution that works is the right one. An LLM (Large Language Model, the AI in the chat window) will happily build you a spaceship when you asked for a bicycle. Tell it explicitly to keep things simple.

2. Modular (split into small building blocks) – many small parts:

Small files, small functions, one file does one thing. Neither you nor the AI understands huge files. Rule of thumb: past 300 lines, split it up.

3. Can I swap it out?

Ask it of every building block: if I want to replace this later, how much do I have to touch? A database, an AI provider or a payment service belongs behind one wrapper, not spread across thirty files.

4. YAGNI (You Aren’t Gonna Need It) – build only what you need now:

No features on stock. No login until somebody has to log in. Every line that does not exist cannot break.

5. Small steps, always runnable:

After every step the project has to start. One step, test, save the state (git commit), next step. Never order ten things at once.

6. You stay in charge:

Read what the AI did. If you do not understand something, ask «explain this in simple words». Code nobody understands is technical debt (work you have to pay back later, expensively) from day one.

📋 Ground rules – always paste these first

You are my developer. I am a beginner. Always follow these rules in this project:

1. KISS (Keep It Simple, Stupid): The simplest solution that works. No extras I did not ask for.
2. Modular: Small files, small functions. One file does one thing. Split anything past ~300 lines.
3. Swappable: Build so that I can replace any building block later. Database, AI provider or external services go behind their own file, not spread across the whole code.
4. YAGNI (You Aren’t Gonna Need It): Build only what I need now. No features on stock.
5. Clear names: Names say what the thing does. Stick to one language consistently.
6. Small steps: Do one step at a time, show me the result, wait for my OK.
7. It must run: After every step I must be able to start the project. Give me the exact command.
8. No secrets in the code: Passwords and API keys (access keys to a service) go into a .env file (a file only for secrets) that is listed in .gitignore (so it never gets uploaded).
9. Explain briefly: After every change, three sentences - what you did, why, what I should test.
10. Ask: If something is unclear, ask a question instead of guessing.

Do two things first:
- Write these rules into an AGENTS.md file in the project folder so they do not get lost.
- Write my idea into an IDEA.md file: what the project does, who it is for and the three most important features. Ask me the questions you need for that first.

🌐 Prompt 1 – Website

I am building a simple website. Follow the ground rules.

Goal: A one-page website for [YOUR TOPIC, e.g. my photo studio].
Content: Hero with title and one sentence, About me, three services, contact with an email link.
Style: [calm and minimal / playful and colourful], large type, lots of white space.

Tech:
- Only HTML, CSS and a little JavaScript. No framework, no build tool.
- Exactly three files: index.html, styles.css, script.js
- Colours, fonts and spacing as CSS variables at the very top, so I can change everything in one place.
- Mobile first, then desktop.
- Placeholder images, I will swap them later.

How to proceed:
1. First show me only the page structure as a list. Wait for my OK.
2. Then build the three files.
3. At the end, tell me how to open the site locally in my browser.

📱 Prompt 2 – Cross-platform app (React Native)

I am building an app for iPhone and Android. Follow the ground rules.

Goal: [YOUR IDEA, e.g. an app that tracks how much water I drink].
Screens: Home with a list, detail screen, settings.

Tech:
- React Native with Expo, TypeScript.
- Navigation with expo-router.
- Store data locally only for now (AsyncStorage). No server, no login, no sign-up yet.
- Folders: app/ for screens, components/ for UI parts, lib/ for logic.
- All storage logic goes into lib/storage.ts so I can swap it for a real database later without touching the screens.
- No state management framework as long as plain React state is enough.

How to proceed:
1. First give me the exact commands to set up the project and run it on my phone with the Expo Go app.
2. Then build screen by screen. Stop after each screen, I will test on my phone.
3. Tell me at every step what I should see on the phone.

🍎 Prompt 3 – iPhone app (Swift / SwiftUI)

I am building an app for iPhone only. Follow the ground rules.

Goal: [YOUR IDEA, e.g. a shopping list].
Screens: List, add entry, detail view.

Tech:
- Swift and SwiftUI, iOS 17 and up.
- One view plus a small view model per screen. No heavy architecture layer.
- Store data with SwiftData, but behind my own store protocol so I can swap the storage later.
- No external packages as long as it works without them.

How to proceed:
1. Tell me step by step how to create the project in Xcode: template, name, settings.
2. For every file, give me the complete content, the file name and where the file belongs.
3. After every step, tell me what I should see in the simulator.

Note: I write in VS Code but build and run in Xcode. Always tell me what happens where.

🧯 Bonus – when something breaks

This is the error:

[PASTE ERROR MESSAGE HERE]

Please:
1. Explain in two sentences what the error means.
2. Name the most likely cause.
3. Make the smallest possible fix. Change nothing else.
4. Do not restructure or rewrite anything without asking me first.

✅ Do this:

Order one thing per prompt.
git commit after every step that works.
Copy error messages in full, do not retype or summarise them.
Drop in screenshots when something looks wrong.
«Explain this to me» is a legitimate prompt.

❌ Skip this:

«Build me Instagram» in one prompt.
Blindly clicking «accept all changes».
Carrying on when it already stopped running.
Writing API keys into the code or pasting them into the chat.
Installing twenty libraries before anything runs at all.

🧭 When you get stuck:

Start a new chat, paste the ground rules again and describe the problem in one sentence. An overlong chat makes the AI confused, not smarter.
And if three attempts lead nowhere: go back to the last working state and make the step smaller.

🍎 Our Setup of Choice:

Currently, Macs are the best and simplest setup for local coding with AI, mainly because of Unified Memory: CPU, GPU, and Neural Engine share the same memory pool. This often makes local models on Mac more straightforward and efficient.

💼 For Professional Local LLM Tasks:

Mac Studio with at least 32 GB memory.

🪟 For Windows-Specific Development:

If you're developing with Unreal Engine, for PlayStation, or generally doing Windows-heavy work, a Windows machine is the better choice.

🧰 Our Devices:

MacBook Pro (Apple M Chip) with 16 GB RAM
Mac Studio (Apple M Chip) with 32 GB RAM
Windows PC with RTX 4070 Super and 64 GB RAM

🦞 Claws:

Claws are personal AI assistants or agent runtimes that don't just chat. They take on real tasks: web, apps, files, terminal, and entire small workflows.

OpenClaw

The most important open-source Claw right now. Originally started by Peter Steinberger. Local, flexible, and clear in the agent and assistant space.

Kimi Claw

Cloud version around OpenClaw. One-click setup, 24/7 online, and significantly easier for beginners.

Hermes Agent

Open-source agent runtime by Nous Research with persistent memory, its own CLI, an API server, and connections to more than 20 chat platforms.

ChatGPT agent

OpenAI's direction for real computer-use and web tasks. Can browse, use files, fill out forms, and execute actions.

Claude Computer Use

Anthropic's computer-use approach. Can see screens and directly control mouse and keyboard.

🧭 What's Happening Right Now:

Established providers are releasing their own personal assistants or programs that can directly control the computer. The direction is clear: less chat, more real execution.

🤖 Names to Know:

OpenAI: ChatGPT agent, formerly Operator
Anthropic: Claude Computer Use
Google: Project Mariner
Kimi: Kimi Claw, Kimi Agent Swarm

🔐 Security:

These tools often get extensive access: browser, files, terminal, logins, clipboard, or entire apps.
Prompt injections are a real problem. A website or file can slip hidden instructions to the agent.
If you run something like this with broad permissions, in the worst case you're voluntarily installing a trojan.
Therefore: only test on machines with little sensitive data, grant few permissions, treat skills and plugins like real code, and never blindly confirm everything.