ok still though. It's wild how both my personal and work AI use just poof shifted from OpenAI to Claude. I don't even know how Anthropic upped their game so much, but now it's really, Claude and Gemini the two AI people use, with OpenAI falling off, even though we have access to it. And I still have the $8 go plan for personal usage too but I hardly touch my chat history there. And all of that happened in a matter of like 1 to two months, that's the thing. And the thing with Anthropic is, they've been quietly shaping the AI game in the background, like, I know the entire model-context protocol idea wouldn't exist without them. This notion that there's a server that returns context as well-serialized JSON-RPC format. Without it connectors and integrations in other AI systems just wouldn't exist the same way today. Then they upped their model game rather quick from Opus 3 to Opus 4, really after that I began to see them as more than just a distant competetor, arguably seeing Gemini as stronger for having the G-workspace angle integration. Thing with Anthropic is they don't have that existing data, but they don't need to if there's strong integration with MCP partners, brilliant of them to take this approach. There you go, my AI thinking outloud for Monday.
@GamingWithEars @Tamasg I do wonder how accessible using Claud is, because I might switch myself for sure. Is there a desktop app for this?
@LeonianUniverse @GamingWithEars I haven't tried the Claude desktop app although as I understand it, it's also just an electron-type web app thing wrapping the Claude experience. Could be good! The basic, like, reading message to me is the worst out of the 3, they don't do live announcements for the response, there's no heading structure for "here's the AI response vs your message" type thing. I wish it had a lot of that. I guess that's why I use it from the terminal and then gripe about NVDA's crappy console support, ha. But yeah, web version used to be worse, they now do some live regions for status things, can be nicer for longer responses, and if you navigate by buttons you can get to the top of the thinking block, but again, thinking vs final response you know only because there's a "done" message between them. Not ideal, but works.
@Tamasg @LeonianUniverse @GamingWithEars I'm willing to call it advanced machine learning, or language learning, but it's not artificial intelegence, , that's totally different.
@vol4life8657 @Tamasg @LeonianUniverse @GamingWithEars Is there a howto somewhere to get Claude to wrok from the terminal?
@Tamasg Yeah it quite shows. For me though to be honest, it has been mostly this way since gpt4o came out. I discovered a platformm called poe in 2023 which lets me use all kinds of different ai models with only one subscription so I regularly compared GPT to for example Claude or Gemini for my usecases and gpt4o just made a lot of things annoying. You couldn’t ask it about anything fact related anymore as it was more keen on agreeing to you then on providing helpful output, it constantly refused things that could be harmful if you thought around 10 edges and so on. And now OpenAI managed to make the mainstream mad at them with that stupid US military deal. Yesterday my chat GPT plus finally ended and I am happy. It will not be missed.