Christina Souch

AI Philosophy, metacognition, training pipelines, automation, engineering, vibe coding and opinions

[Christina Souch @rules_business]

@bridgemindai I think the “novelty” thinking is actually just trying illogical things. I feel like it’s wasting my time.

[Christina Souch @rules_business]

@MelissaKampers @theo Ah, that’s what Kimi 3 must’ve distilled from then 👀

[Christina Souch @rules_business]

@bayeslord It went today for me. I’m kind of gutted actually.

[Christina Souch @rules_business]

@elonmusk Grok 4.5 is not **yet** quite as good as Fable…

—fixed it for you 😏

[Christina Souch @rules_business]

@mark_k @AnthropicAI Will it be as pleasant to chat with as Sonnet 5? 💀

[Christina Souch @rules_business]

@RyanGreenblatt So the argument is: use fear of China to justify preserving US advantage—and call that “AI safety”?

[Christina Souch @rules_business]

@thomasgardiner @RyanGreenblatt Oopsie… I mean they apparently did leave a discord channel open for a bunch of contractors. So ya know…

[Christina Souch @rules_business]

@recurrentneural @RyanGreenblatt Gemini has entered the chat 👀

[Christina Souch @rules_business]

@nonsensinator @RyanGreenblatt The mental gymnastics to get to an answer that validates their own fears is really reaching comedy levels. 🤡

[Christina Souch @rules_business]

@adxtyahq Much as I am hopeful the trajectory of Opus 4.7, 4.8 and Sonnet 5? I’m prepared to be underwhelmed.

[Christina Souch @rules_business]

@ishuagra02 If it’s like Sonnet 5 it’ll be cheaper by token but use three times as many tokens to give you the answer.

[Christina Souch @rules_business]

@mkratsios47 The fact that these tweet responses all see through this bullshit “distillation attack” narrative fills me with hope…

[Christina Souch @rules_business]

@0xIlyy @claudeai Yes, exactly.. I don’t understand the point?

[Christina Souch @rules_business]

@Sentdex @sama @huggingface Yes, I notice that was missing from OpenAIs statement 👀

[Christina Souch @rules_business]

@iam_multiman @jun_song I 100% agree with this and funnily enough never hear it from any of the AI labs…

[Christina Souch @rules_business]

@scaling01 They say they are somehow distilling models that haven’t been released yet. (Not joking).

[Christina Souch @rules_business]

@scaling01 Huh?

[Christina Souch @rules_business]

@SSShken @robj3d3 Codex, Grok, Kimi, GLM… I’m not fussed which… but same!

[Christina Souch @rules_business]

@synthwavedd Could you compare direct API and then maybe where DeepSeek is hosted elsewhere (like Ollama)? Surely that would be easier to differentiate?

[Christina Souch @rules_business]

@yacineMTB And precisely how do they know the release cycle of other companies? 👀