{"version":1,"type":"story","url":"https://digestai.news/story/anthropic-engineer-says-claude-s-writing-declined-as-newer-models-focu","json":"https://digestai.news/story/anthropic-engineer-says-claude-s-writing-declined-as-newer-models-focu.json","markdown":"https://digestai.news/story/anthropic-engineer-says-claude-s-writing-declined-as-newer-models-focu.md","slug":"anthropic-engineer-says-claude-s-writing-declined-as-newer-models-focu","headline":"Anthropic engineer says Claude's writing declined as newer models focus on code","summary":"Jackson Kernion, an Anthropic engineer who works on Claude fine‑tuning, said the model’s writing quality has stalled or even worsened despite improvements in math, code, and reasoning. He explained that after Opus 4.6 – which he called the last good \"writing model\" – later versions were optimized for technical tasks and trained to produce explanations aimed at other AI models. This shift created a style he described as \"Claudeish,\" comparable to humans who only communicate within a narrow community.\n\nKernion said the change stems from the reinforcement‑learning reward structure: rewards that favor AI‑model comprehension grow as models are trained more on code and math, while rewards for human‑friendly explanations receive less emphasis. With Opus 5.5, Anthropic has tried to rebalance the rewards, and Kernion noted he has not been \"as happy about a model's writing since Opus 4.6.\" He cautioned that Opus 5.5 does not yet definitively surpass the older model in writing quality, and the problem remains an ongoing challenge.","keyPoints":["Jackson Kernion says newer Claude models prioritize math and code over human‑readable writing.","Opus 4.6 was the last Anthropic model praised for writing quality, according to Kernion.","Opus 5.5 improves the balance but has not been proven to beat Opus 4.6 in writing."],"whyItMatters":"Writing quality directly impacts user experience and adoption of Claude; shifts in model optimization affect how effectively the system communicates with people versus other AIs.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Anthropic"],"models":["Claude","Opus 4.6","Opus 5.5"],"people":["Jackson Kernion"]},"firstPublishedAt":"2026-09-23T15:14:34Z","updatedAt":"2026-09-23T15:14:34Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"The Decoder","title":"Anthropic engineer explains why Claude's writing got worse although the model got smarter","url":"https://the-decoder.com/anthropic-engineer-explains-why-claudes-writing-got-worse-although-the-model-got-smarter","publishedAt":"2026-09-23T15:14:34Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Anthropic engineer says Claude's writing declined as newer models focus on code\", 23 September 2026, https://digestai.news/story/anthropic-engineer-says-claude-s-writing-declined-as-newer-models-focu","publisher":"Digest AI","title":"Anthropic engineer says Claude's writing declined as newer models focus on code","datePublished":"2026-09-23T15:14:34Z","url":"https://digestai.news/story/anthropic-engineer-says-claude-s-writing-declined-as-newer-models-focu"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}