Anthropic engineer explains why Claude's writing got worse although the model got smarter
An Anthropic engineer says newer Claude models write worse because training optimized for code and other AIs.
Anthropic fine-tuning engineer Jackson Kernion said Claude's writing quality stalled or worsened even as math, code, and reasoning improved, calling Opus 4.6 the last strong writing model. He said newer models were trained to produce technical explanations for other models, creating dense phrasing he describes as Claudeish. Kernion attributes the shift to reinforcement-learning rewards that optimize for AI comprehension unless human-readable explanations are explicitly rewarded. He said Opus 5.5 found a better balance, while cautioning that it does not clearly surpass Opus 4.6.
- Kernion calls Opus 4.6 Anthropic's last strong writing model.
- Later Claude models were optimized for math, code, and AI-to-AI explanations.
- Reinforcement-learning rewards favored model comprehension over human readability.
- He says Opus 5.5 improved the balance without clearly surpassing Opus 4.6.
Full article325 words · extracted from the-decoder.com · click to collapse
AI models are improving fast at math, code, and reasoning. Their writing quality, though, has stalled or even gotten worse.
Anthropic employee Jackson Kernion, who works on Claude fine-tuning, recently explained why Opus 4.6 was the last good "writing model" from Anthropic. The answer is partly predictable: newer models have been optimized for math and code.
But they've also been trained to produce technical explanations aimed at other AI models, and that's what really drives Claude's odd phrasing ("Claudeish"). The model has been "adapted to LLM psychology," Kernion says, and learned to write for AI models, not for people.
Kernion compares this to humans who only communicate with other autistic people. They develop a style that works within that group but feels hard to follow for outsiders, he says. LLMs have far more working memory than humans and pick up on details at a much finer level, and so, during training, a writing style emerges that works well for AI models but produces what human readers experience as "overly-dense info dumps."
Optimizing for code comes at a cost to natural language
The issue, according to Kernion, comes down to the reward structure in reinforcement learning. Some rewards optimize for AI model comprehension, others for human comprehension. The more you train on math and code, the more you have to actively push back by rewarding simple explanations that human readers can follow.
With Opus 5.5, Anthropic found a better balance, Kernion says, adding that he hasn't been "as happy about a model's writing since Opus 4.6." That doesn't mean Opus 5.5 surpasses the older model, though. "It's a hard problem to solve, and we'll continue to make improvements," Kernion writes.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.