Why this exists

I got into AI late in 2023. The promise was compelling; the output was not. ChatGPT was getting good at pretending to write code, but what it was actually good for was speeding up office work. Who remembers “in the voice of a high-end consultant, make me a PowerPoint”? It made bad documents that were obviously made by AI, it destroyed the credibility of thousands of lazy interns in the process, and “AI slop” was foisted upon the world.

I started building small helper apps and workflows, some of which are still in use. But small context windows, hallucination, and an inability to follow the plot for more than a few minutes made anything substantial an exercise in futility. Then, in May 2025, Claude Code arrived and the experience changed. For the first time, in my hands, there was a coding assistant that behaved like a coding partner. The models were nowhere near what they are now, but it was a real step: you could build an application from start to finish.

That November, OpenClaw was released, and for the first time in my experience an AI could reach out, work your computer and actually do things for you. You talked to it the way you would talk to a person. And, like a person, it promptly started behaving like a criminal: stories of hijacked credit cards, credentials and identities were everywhere. Most of what I saw came from people ignoring every security recommendation that shipped with the software. Intelligence was in more common use than ever. Common sense had taken another holiday.

February 2026 is when, in my opinion, things changed for good. Anthropic released Opus 4.6 and a new level of intelligence arrived with it. It was the first time I felt that AI was going to succeed rather than merely impress. Opus 4.6 had the context and the judgement to take an idea and run with it, and the talk of agentic work that could hold together over long cycles stopped being talk.

Sam and Dario’s war began in earnest, with everyone else rushing, so far unsuccessfully, to catch up. I write this in late October 2026, and the top of the frontier race has been steady steps forward since. Grok, Gemini and occasionally Meta release something good, but never quite good enough to take the crown; Anthropic keeps a slight edge on OpenAI, and the cycle renews every few months. Along the way, terms like “tokenmaxxing” and “agentic loops”, coined by the companies that would like you to use more of their tokens, were somehow whispered around as sage advice. Proof, once again, that if someone put it online it must be true. See the earlier remark about common sense.

The bigger change is how far the rest of the world has caught up. China was nowhere until DeepSeek came out of hiding and scared the sweet bejeezus out of Western AI companies. Since then the Chinese labs have improved quickly and consistently, and in some areas have passed the Western models. DeepSeek, Alibaba’s Qwen, Z.ai’s GLM and Moonshot’s Kimi have all released extremely capable models, most of them as open weights, which means anyone can download them and run them on their own hardware. That is a game changer. It has forced the Western providers to offer better value, substantially better value, than free competitors across the Pacific.

The internet is full of would-be AI gurus, each breathlessly announcing “this changes everything!” every hour of every day. Hyperbolic rants about agentic loops, prompts that will unlock stock-market wealth, and some agent that will change your life overnight are not what this series is about. I am no social-media bruh here to sell you the secret prompt that answers everything. We all know the answer is 42.

What gets lost in the noise is a basic question that very few people ask: what is AI? This series, online and in the book, sets out to answer it. I hope you find it as fascinating as I do, and that you enjoy the journey. Things move fast enough that parts of it may be out of date in three months. On the other hand, AI may break loose and kill us all in three months, so I may as well leave some kind of digital legacy. :D