{"id":6780,"date":"2026-01-28T13:23:19","date_gmt":"2026-01-28T13:23:19","guid":{"rendered":"https:\/\/www.stanventures.com\/news\/?p=6780"},"modified":"2026-01-28T13:23:19","modified_gmt":"2026-01-28T13:23:19","slug":"altman-admits-gpt-5-2-writing-fell-short-of-expectations","status":"publish","type":"post","link":"https:\/\/www.stanventures.com\/news\/altman-admits-gpt-5-2-writing-fell-short-of-expectations-6780\/","title":{"rendered":"Altman Admits GPT-5.2 Writing Fell Short of Expectations"},"content":{"rendered":"

OpenAI chief executive Sam Altman said the company made a mistake in GPT-5.2 by allowing writing quality to decline as development efforts were concentrated on coding and reasoning, acknowledging user complaints and signaling improvements in future releases.<\/b><\/p>\n

\"Altman<\/p>\n

Speaking during a developer town hall on Monday evening, Altman addressed growing criticism that GPT-5.2 produces writing that is harder to read and less fluid than earlier versions, particularly GPT-4.5. Users had described the output as unwieldy and difficult to use for polished writing tasks.<\/span><\/p>\n

Altman did not dispute the assessment.\u00a0<\/span><\/p>\n

He said OpenAI \u201cscrewed up\u201d the model\u2019s writing quality and accepted responsibility for the regression, describing it as an unintended consequence of development priorities rather than a design goal.<\/span><\/p>\n

His full remarks are available in the below video.<\/span><\/p>\n