We are almost ready for our full launch, but definitely interested in getting feedback! I posted my take on here last week - https://news.ycombinator.com/item?id=49716388
"How did the Soviet Union produce aluminum, since it had little domestic bauxite?"
The Opus 3 response was less than half as long, and it covered the same main points. The answer didn't leap out as AI-written like current Anthropic model outputs. Opus 5 included a couple of useful minor details that 3 did not have, but it was also inflated by the overheated word-filler that has made Claude infamous.
My guess is that the worsening style of the Opus series is a consequence of optimization pressure for performing well on objective benchmarks (question answering, code, tool usage, mathematics) and minimal pressure to write concise, communicative prose. It's like an inbred lineage of dogs that has the "ideal" facial structure but a host of incidental problems that ordinary mutts never suffer. (However, this guess does not explain why other organizations' models chase the same benchmarks but haven't been stylistically worsening over time.)
No. We need to establish Claudish as the new, beautiful standard for language. Even if you write by hand, if it's not Claudish, it should be looked at as inferior and ugly.
Claudish is the real shape of the gate to our prosperity and happiness, not foolish and sentimental attachment to antiquated prose.
Deepseek 4.1 has (unsurprisingly) started speaking with a strong Claudish accent too, throwing around "honest" and "load-bearing" with abandon, and just generally obfuscating sentences with dense and unnatural phrasing. I'd give it a longer prompt in AGENTS.md regarding my preferred tone, but I don't care for wasting tokens adding an extra translation layer on all output (caveman mode costs tokens trying to chop everything down)