Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

AI Coding

  • TBD: How to pick the model?
  • Explain the @ commands and the / commands

LLMs - Large Language Models

  • autocomplete

  • context

  • tokens

    • a word
    • part of a word
    • a single letter

  • limitations

    • No real intelligence just patterns and probabilities
    • Hallucinations

  • prompts, prompt engineering

    • clear and precise
    • provides context
    • iterated and refined

  • Ask only one thing in every prompt.

General purpose LLMs

Where you can chat, create images, create music etc.

Models

by Antrhopic

  • Claude Opus (4.x series)

  • Claude Sonnet (4.x series)

  • OpenAI

    • Models
      • GPT-5.5
      • Codex
      • …
      • DALL-E 2
      • Sora

by Google

  • Gemini

by Microsoft

  • Phi (but it is not very popular)

Harnesses

  • GitHub co-pilot (Microsoft)

    • In VS Code
    • CLI
    • GitHub UI to chat
    • GitHub Issues - assign to copilot
    • Available Models (from other vendors)
      • GPT-5.4
      • GPT-5.3-Codex
      • GPT-5.2
      • GPT-5.4 mini
      • GPT-5 mini
      • GPT-4.1
      • Claude Sonnet 4.6
      • Claude Haiku 4.5

  • Cursor

    • A modified version of VS Code. An independent startup.

  • Codex (OpenAI)

    • In VS Code
    • CLI

  • Gemini Code Assist (Google)

    • In VS Code
    • Gemini CLI is deprecated
    • Antigravity CLI

  • Claude Code (Anthropic)

    • CLI - It is a command line tool.
    • It can work as an extension in major IDEs.

  • Windsurf (Codeium)

    • Stand alone editor
    • Plugin to JetBrains, VS Code etc.

  • Tabnine

GitHub Co-pilot

  • Sign up to co-pilot

  • Install VS Code

  • Install the co-pilot plugin

  • Authenticate

  • Inline Chat

  • In the chat window ask for some code and show the various buttons.

In Docker container

In order to allow the CLI tools to roam freely on the code-base without the fear of them making changes in other places of the disk or the operating system we can use them inside Docker containers.

See my Rust, Python, and Perl containers.

Antigravity (in place of Gemini)

Add the following lines to your Dockerfile:

RUN echo Install Antigravity   && \
    curl -fsSL https://antigravity.google/cli/install.sh | bash  && \
    echo done

Add the following line to the .bashrc file of the container.

export PATH="/home/ubuntu/.local/bin:$PATH"

When you start Antigravity inside the container you’ll need to authenticate. It shows you a link that when I tried was brokent. It inserted spaces in the URL. So I had to copy the URL, remove the spaces and only then I could use it.

GitHub Co-pilot CLI

RUN echo Install GitHub co-pilot CLI   && \
    curl -fsSL https://gh.io/copilot-install | bash  && \
    echo done

Codex CLI

apt install -y gawk

RUN echo Install Codex   && \
    curl -fsSL https://chatgpt.com/codex/install.sh | CODEX_NON_INTERACTIVE=1 sh && \
    echo done

MCP - Model Context Protocol

MCP = Model Context Protocol (an open AI standard)

Collection

  • GitHub Copilot by GitHub / Microsoft

    • CLI
    • In VS code
    • On GitHub (assigning issues to it)

  • Google Antigravity by Google

  • Codex (AI agent) by OpenAI

    • CLI

  • Ollama

  • Open Code Go

    • Models such as KimiK and DeepSeek.
    • OpenCode with VS Code, with the OpenChamber plugin which provides a Copilot-like experience.

  • DeepSeek

  • Claude Code

  • Windsurf with SWE1.5, SWE1.6 (now called Devin?)

    • VS code plugin

  • VS Code with Continue and OpenRouter models Deepseek v4 Flash and MoonshotAI Kimi K2.6 (free)

  • Meta AI and Google AI to ask programming questions.

  • AI on premise via huggingface should be free in general for private use. https://huggingface.co/docs/hub/billing