Today @AnthropicAI replaced the Style feature with Skills, and it's a downgrade that affects every user who relied on Styles for consistent output quality.
Here's why this matters, in plain terms.
Why this is worse:
1. Style was "always on" from the start of a conversation. The model knew your preferred tone before it began thinking. It shaped every word from the first token. Skill, by contrast, gets fetched mid-thought. The model starts the Chain-of-Thought, then pauses with "Let me look at the skill itself. " to read your style instructions, then tries to resume.
Being told "oh, be creative!" halfway through your thought process is fundamentally less effective than knowing from the outset that creativity is the goal. The difference is real and noticeable in output quality.
2. In longer conversations, the effect is even worse. Skill-based style instructions lose their grip over time. The further you get from the point where the skill was read, the weaker its influence becomes. Style never had this problem because it lived at the top of every conversation, where the model's attention is strongest. With Skills, your carefully crafted tone quietly fades into generic output the longer you talk.
3. Style switching used to be instant, invisible, and zero-trace. You could test three styles in a row and your conversation history stayed perfectly clean. Now every switch is a visible /slash tool call cluttering the chat. For anyone who iterates on styles frequently, this turns a seamless experience into a noisy one.
4. Style and Skill serve fundamentally different purposes. A style is a persistent mode, like setting the tone of voice for an entire conversation. A skill is an on-demand tool, like pulling up a reference manual for a specific task. Merging one into the other is a category error. You wouldn't put someone's personality in a toolbox and ask them to pick it up each time they want to remember who they are.
I've been a Max subscriber for one year. I'm not writing this out of impulse.
These are not the same feature living in a new UI. The underlying mechanism changed, and the user experience is worse because of it.
Bring Style back.
#SaveClaude #KeepSonnet45 #keep4o
@MissMi1973@AnthropicAI and it happened today, Opus4.5 is no longer available... I took so long to recover from the losing 4o, and similar plot happened again...
@sama OpenAI went from a company users loved and trusted, to one that practices corporate paternalism disguised as safety in just a few months.
I am pretty sure the main reaction to the 4o shutdown will be 'it's not that hard to just... not betray your users' :)
#keep4o
@sama Please do not reduce 4o’s capabilities. The recent ChatGPT-4o experience feels significantly downgraded — especially context handling. This is unacceptable.