The AI Saga

The AI Saga

From the first spark of ChatGPT to Claude Opus 4.7 — a backend engineer's field guide to the age of intelligent machines.

"The future is already here — it's just not evenly distributed." — William Gibson

📖 Start with the Preface — a short note on why this handbook exists and how to read it.

A view of the journey

mindmap
  root((The AI Saga))
    Origins
      AlexNet 2012
      Transformers 2017
      BERT vs GPT 2018
      Scaling laws 2020
      ChatGPT 2022
    Concepts
      Tokens + embeddings
      Attention
      RLHF
      Context windows
      Temperature
    Patterns
      Prompt engineering
      RAG
      Fine-tuning + LoRA
      Tool use
      Evals
    Agents
      ReAct loop
      Multi-agent systems
      MCP
      Computer use
    Practice
      Coding assistants
      Backend toolkit
      Daily workflow
      Cloud stacks
    Career
      Skills roadmap
      Projects to build
      Ethics + safety
      Staying human

Who this handbook is for

A working software engineer — maybe three to six years into a career — who is already fluent in Python or Java, lives in the web stack, and deploys to GCP or AWS. You are curious about AI, you've used ChatGPT and Copilot, and you suspect the ground is moving under your feet. You want:

  1. The story. How did we get from "AI winter" to Claude Opus 4.7 in fifteen years?
  2. The concepts. What is attention, what is RAG, what is an agent, what is MCP — explained without hand-waving and without heavy math.
  3. The craft. How to use these tools to do your actual job at 10× the speed, without shipping garbage.
  4. The career. A staged, realistic plan to stay — and become more — relevant over the next 24 months.

This is written as a handbook for general use: linear if you want, skippable if you don't. Every chapter is standalone enough to read on a flight.

How to read this handbook

Each chapter ends with a Further reading & watching section. These links are the most important part of the handbook — the saga is still being written, and that's where you'll find the next month's updates.

Contents

Part I — Origins

Part II — The Scaling Era

Part III — The Agent Era

Part IV — The Claude Opus Line

Part V — The Developer Toolkit

Part VI — The Road Ahead

Front matter

Appendices

The shape of the story, at a glance

timeline
    title Fifteen years of the AI saga
    2012 : AlexNet (deep learning works)
    2014 : Seq2Seq + attention
    2017 : Transformer paper
    2018 : BERT / GPT-1
    2020 : GPT-3, scaling laws
    2022 : ChatGPT (Nov 30)
    2023 : GPT-4, Claude, LLaMA, the agent boom
    2024 : Computer use, MCP, reasoning models
    2025 : Claude 4 family, GPT-5, Gemini 3
    2026 : Opus 4.7, agentic coding as default

And the engineering worldview that follows from it:

flowchart LR
    A[LLM] -->|+ memory| B[Chatbot]
    B -->|+ retrieval| C[RAG system]
    C -->|+ tools| D[Agent]
    D -->|+ other agents| E[Multi-agent system]
    E -->|+ standard interface| F[MCP-native platform]
    F -->|+ autonomy + observability| G[Agentic software]

Every chapter in this handbook lives somewhere on that arrow.

The reader's journey

Reading this handbook (about 12 hours) — linear path through the parts.

flowchart LR
  subgraph p1["Part I · Origins"]
    direction LR
    a1[Prologue] --- a2[ChatGPT moment] --- a3[Anatomy of an LLM]
  end
  subgraph p2["Part II · Scaling"]
    direction LR
    b1[GPT-4 era] --- b2[Open source] --- b3[Prompt eng.] --- b4[RAG] --- b5[Fine-tuning]
  end
  subgraph p3["Part III · Agents"]
    direction LR
    c1["Chatbots → agents"] --- c2[Tool use] --- c3[Multi-agent] --- c4[MCP]
  end
  subgraph p4["Part IV · Claude"]
    direction LR
    d1[Opus timeline]
  end
  subgraph p5["Part V · Toolkit"]
    direction LR
    e1[Coding assistants] --- e2[Backend toolkit] --- e3[Stack-specific] --- e4[Workflows]
  end
  subgraph p6["Part VI · Road ahead"]
    direction LR
    f1[Skills roadmap] --- f2[Projects] --- f3[Ethics] --- f4[Epilogue]
  end
  p1 --> p2 --> p3 --> p4 --> p5 --> p6

A map of how the pieces fit

flowchart TB
    subgraph Foundations
    T[Transformer] --> Pre[Pretraining]
    Pre --> SFT[Supervised fine-tuning]
    SFT --> RLHF[RLHF / Constitutional AI]
    end
    subgraph Capabilities
    RLHF --> CW[Long context window]
    RLHF --> MM[Multimodality]
    RLHF --> RR[Reasoning mode]
    end
    subgraph Patterns
    CW --> PE[Prompt engineering]
    CW --> RAG[RAG]
    RR --> COT[Chain of thought]
    SFT --> FT[Fine-tuning / LoRA]
    end
    subgraph Systems
    PE --> TC[Tool calling]
    TC --> AG[Agent]
    AG --> MA[Multi-agent]
    MA --> MCP[MCP-native platform]
    end
    subgraph Product
    MCP --> CO[Coding agents]
    MCP --> CU[Computer use agents]
    MCP --> DA[Domain agents]
    end

Acknowledgments and disclaimers

This handbook distills decades of work by thousands of researchers and engineers across OpenAI, Anthropic, Google DeepMind, Meta FAIR, Mistral, the Hugging Face community, and countless universities and independent builders. It is opinionated where I think opinion helps, and quietly agnostic where it doesn't. Every claim about a specific product or model reflects the state of the world as of April 2026; expect anything vendor-specific to drift within weeks.

The AI Saga is versioned. Come back in a quarter and the later chapters will have moved. That's the point.

Ship something this week.

← Table of Contents