WHY GPT-5.2 FEELS DIFFERENT

So 'they' were only a couple of days out. GPT-5.2 landed yesterday. And was the feedback heard? Well, it turned up in pressed trousers and a lanyard.

Given Brightbeam's early use, the press release - and far deeper reports from those with pre-access - OpenAI’s intent seems clear.

This is a model shaped for professional output: structured documents, spreadsheets that survive scrutiny, slides that do not need rescuing, long-context reasoning and multi-step executions.

The emphasis has shifted from creativity to utility, from cleverness to economic usefulness.

And on those terms, it seems to succeed. So far at least.

Complex tables hold together. Financial logic follows through to the end. Planning artefacts look coherent rather than improvised. Long contexts degrade less and hallucinations retreat, which matters when decisions, not demos, are on the line.

But. And it'll likely be a big old but for many: Something has been traded away.

5.1 Thinking had a peculiar temperament. It could wander, overthink and occasionally ramble. But it also had a habit of stumbling into ideas you did not explicitly ask for. It behaved like a clever but undisciplined collaborator: unreliable in formal settings, oddly fertile in exploratory ones. You would not trust it with the final draft. You could trust it to unsettle your assumptions.

5.2 is different. It is tidier, more bounded, more intentional. The edges have been filed down. The output is cleaner and the tone more restrained. You get fewer surprises, but also fewer moments of genuine creative friction. It feels less like a sparring partner and more like a senior analyst who reliably delivers exactly what was requested.

So yes, welcome GPT-5.2. It does the serious work with fewer errors and less supervision. But there will be a quiet nostalgia for 5.1’s quirks, for the slightly unruly collaborator. The one who sometimes produced the only thought you had not circled on the whiteboard.

The image above was the first produced by ChatGPT-5.2 when asked to illustrate this post.