o1 and Sora: Reasoning and Reality

December has arrived, and with it two OpenAI releases that feel less like product launches and more like punctuation marks on the strangest, most accelerated year of my life. The full o1 model shipped, and Sora launched to the public. I have been sitting with both, and I want to reflect before the year ends and the blur hardens into hindsight.

o1: The Reasoning Breakthrough

OpenAI previewed o1 in September under the codename "Strawberry," and the full version arrived in December. It is unlike any model I have used, and I want to be careful describing it because the standard vocabulary undersells it.

o1 does not just answer. It thinks. Before responding, it runs an internal chain-of-thought — a hidden scratchpad where it breaks the problem into steps, considers approaches, catches its own errors, and revises. You see the final answer; you do not see the deliberation, though OpenAI exposes a summary of it. The result is a model that solves problems no previous language model could touch: competition mathematics, advanced physics, complex multi-step reasoning where the answer is not retrievable from training data but must be derived.

I gave it a problem I have used for years to humble confident models: a combinatorics puzzle involving seating arrangements that requires keeping track of five interacting constraints simultaneously. Every prior model I have tried — GPT-4, Claude, Gemini — has failed it, usually confidently and incorrectly. o1 solved it. Slowly, deliberately, correctly.

The implication lands harder than the benchmark. If a model can reason rather than retrieve, the space of problems AI can genuinely help with expands enormously. Real intellectual work — research, analysis, design, the kind of open-ended problem-solving that defines serious creative and technical practice — is not about recall. It is about working through uncertainty. o1 is the first model that does that in a way that feels like a collaborator rather than a search engine.

Sora: The Public Launch

And then Sora. The video generation model that captivated me back in June launched to the public in December as a standalone product at sora.com, with a timeline-based editor, the ability to remix and extend generations, and a community feed of creations.

I have been back in it every evening. The quality has improved since my early access — sharper motion, better consistency, fewer of the surreal dissolution failures I wrote about in June. I generated a sequence this week of a slow camera move through a Monet-style garden at golden hour, the light catching on a pond, willows shifting in a breeze I had specified. It is the kind of thing I would have spent a month painting. It took ninety seconds.

I keep returning to the phrase I used in June: painting with time. Sora is the first medium where an artist manipulates duration directly. Six months in, that still feels right, and the tooling has caught up to the concept.

Two Threads, One Year

Here is what strikes me, holding these two releases side by side.

o1 extends AI's reach into reasoning — the capacity to work through hard problems in domains where answers must be derived, not looked up. Sora extends AI's reach into reality — the capacity to render the world, to imagine it in motion, to externalise perception into shareable artefacts. One thinks. One sees. Together they sketch the outline of an intelligence that can do both.

I have spent the year collecting building blocks — speed, persistence, memory, frontier reasoning, perception, action, generation. Looking at the pile now, in December, I notice that every major capability I would need to build something genuinely ambitious is no longer hypothetical. It exists. It is imperfect, uneven, early. But it exists.

That is the quiet revelation of this week. The question is no longer can these capabilities be built. The question is what do we build with them.

I have spent evenings this month walking in the cold, turning that question over. I do not have the answer yet. But I have the materials, and I have the year behind me, and the fire is lit.

Live curiously and give generously.