@arlotests@SemiAnalysis_ Just read the article and stop asking questions for something they have already given you in the article.
But yes, they do apply to Opus. The 5x number actually does apply to Opus vs Sol.
@brave Hey guys. Have you realized that your browser is not parsing strings on the AI-generated answers? They remain in whatever format the model generated rather than becoming human-readable.
@bungehustleke@HedgieMarkets This is a braindead assertion. It waters down the brutality of real colonialism. Also throws out any real nuance in conversations about fair use, fair burden, crawling courtesy, and transformative processing and replaces it with a hyperbolic assertion. Do better fellow Kenyan.
@PtrPomorski@karpathy I don't think so. I think Anthropic is the one that had a shot. For any lab, getting Karpathy is the money shot, not Karpathy having a chance at any lab...
@3blue1brown Hey Grant. This channel has cloned your voice on YouTube and has been posting for the past month or so. It's exactly your voice. I have reported it and I think you should too because your voice will reach much further: https://t.co/E1PtZw5qQJ
@ntbrown01@davepl1968 And that's how I know you have only used models as chatbots like ChatGPT and don't understand any internals. Internally, they are fully deterministic, it's just that random sampling provides more interesting answers!
@EckertAnthony@davepl1968 Actually, no! If you gave a model the same input twice and chose the token the model has the most confident answer in, it will give you the same answer. The issue is that there is random sampling among tokens to prevent boring responses and repetition!
@SnazzyLabs How about https://t.co/qsDYT65uMn? Catchy!
And how about that Macbook Neo review? I pay you zero dollars, except likes and comments, but I demand that review be a documentary, 50 minutes plus, given how long it's taking. It better touch on how NeXT's WebObjects led to the Neo...
@AlgorithmicSimp I am glad to hear that you are okay. Thanks for replying. All the best in your endevours and we will still be here when you come back.
@AlgorithmicSimp Hey mate! You have been missing in action for a whole year now. Are you okay?
I hope you are. I'm tweeting at you because there was no other way for me to reach you. I am now watching a number of your videos for nth time because I'm new to ML. Cheers mate!
@automate_archit@YahooFinance "Claude Code is basically an autonomous agent framework disguised as a CLI tool"
I don't know mate, who thought it was just a CLI Tool? You might be the only one who thought that.
Second, it's only semi-automomous at best.
@yashhsm @Fried_rice @claudeai What kind of logic is this?! 1. They have already sent about 10K DMCA takedowns, that ship has sailed 2. Nobody would be fooled 3. One of their pillars is keeping their AI as closed-source as possible, that would be a big about turn that would again NOT FOOL ANYBODY!!!
@whoisheartbreak@SpaceX Wow! Hey @whoisheartbreak, congratulations! I think it has been quite some time since we last saw you on the launch hosting role. Anyway, it must be an amazing job for you to stay that long. All the best to you though.