Larry Ellison said something precise in his recent presentation. He didn't dramatize it. He said this: "A lot of the code Oracle is writing — Oracle isn't writing. Our AI models are writing." It's a technically specific statement about how software development works inside Oracle today. That precision invites serious analysis, without panic or automatic applause.

The nuance matters. Ellison isn't talking about experiments or assistants that suggest code for human review. He's referring to models generating production code. That's a qualitative shift. In significant parts of Oracle, software writing no longer depends on a human at the helm. The infrastructure has absorbed it.

What he didn't mention in that speech is just as revealing: Oracle is cutting between twenty and thirty thousand jobs, up to eighteen percent of its global workforce. Not because of financial trouble. It just reported its strongest revenue quarter in fifteen years. These layoffs are funding the expansion of AI data centers. It's the largest tech layoff of the year, and it didn't come up in the presentation celebrating AI's role in writing code.

Ellison rejects the idea of a "software apocalypse." He argues that AI expands software rather than destroying it, generating more applications and more infrastructure demand. He may well be right. Software won't disappear. But that expansion is happening with fewer human programmers and more control concentrated inside Oracle. Value moves upward; it doesn't get distributed. That distinction is left out of the speech.

I recognize this dynamic from other historical contexts. Gutenberg's printing press didn't reduce text production; it multiplied it. Yet control over publishing, distribution, and profits concentrated in the hands of press owners and networks. Scribes didn't disappear immediately; their function was absorbed into the machine and into whoever operated it. What changed wasn't the amount of text. It was who profited from it.

AI generating production code follows that same line. It doesn't democratize software. It industrializes it under the control of whoever owns the models, the infrastructure, and the data centers. Ellison isn't eliminating the programmer. He's internalizing that function within Oracle. It sounds technical, but it implies structural changes: knowledge once scattered across thousands of engineers now resides in proprietary models. If an engineer leaves, the company loses something unique. If the model learns, Oracle keeps everything.

The history of tools is consistent on this point. The iron plow in the Middle Ages didn't end farming; it reconfigured who controlled land and production. AI is doing something analogous with code. These shifts don't have a fixed destination: they depend on the structures surrounding them.

The operative question for any tech company is direct: if AI writes production code while you cut thirty thousand jobs in a record revenue year, what does that say about the future of work in tech? The answer isn't in the presentation. It's in the numbers, which point toward concentration rather than expansion.

This isn't happening in isolation. Recently, Palantir's Alex Karp defined who has a future in the tech economy. Nvidia's Jensen Huang proclaimed that AGI already exists. Ellison revealed that AI writes Oracle's code. Three CEOs, different approaches, but convergence in one direction: the power to decide what work matters, what infrastructure supports it, and who controls the models keeps accumulating in fewer hands. It's not a conspiracy. These are rational incentives within a structure that rewards centralization.

What's interesting is that Ellison gets something important right. Software will expand. Demand for intelligent systems will grow. The problem isn't that AI writes code. It's about governance: how that code gets written, deployed, and monetized. There are communities and researchers building open models and cooperative infrastructure, where control doesn't depend on a single company. Those alternatives exist, even if they lack the shine of an Oracle keynote.

The history of technologies shows that outcomes depend on institutional context, not just the tool itself. The printing press concentrated power in some places and dispersed it in others, depending on the laws and structures around it. The internet started decentralized and ended up dominated by platforms, through economic decisions, not through the technology itself. Generative AI stands at that same crossroads.

I'm not sure how this gets resolved at scale. It's more complex than it looks. But naming what's happening with precision is the first step. Ellison was exact about the technology. He left out the human implications. Putting both together isn't pessimism. It's the foundation for real conversations about work in tech and who shapes it.

Stones don't lie, but historians sometimes do.

Sources:

1. TechFinitive — Larry Ellison's statement on AI generating Oracle's code

2. Medium — Analysis of Oracle's 20,000-30,000 job cuts and their connection to AI data center investment

3. Oracle Q4 2025 financial reports — Strongest revenue quarter in 15 years