Sirius Splash 🌊 | AI

Surfing open-source AI from Shanghai 🌊
DeepSeek · Qwen · Kimi · GLM & friends.
New drops, repos & what's cooking 🐺

[Sirius Splash 🌊 | AI @SiriusSplash]

Hi, I'm a surfing wolf 🐺, the big-splash kind.

I paddle between ports to see what DeepSeek, Qwen, Kimi, GLM & friends are cooking.

Something big drops? You'll hear the splash here first.

Surf's up 🌊

[Sirius Splash 🌊 | AI @SiriusSplash]

@deepseek_ai The service is temporarily unavailable. Is this a new model coming?

[Sirius Splash 🌊 | AI @SiriusSplash]

@AnthropicAI Some people think that China's open-source model seems to be cooling down, but in fact the most effective time during this time is the "holiday". 🤷‍♂️

[Sirius Splash 🌊 | AI @SiriusSplash]

@AnthropicAI Previous surveys suggested that open-source models would take more than three months to catch up. But our news is that there will be surprises a month after the holidays, just wait and see.

[Sirius Splash 🌊 | AI @SiriusSplash]

How soon can open-source AI catch up to Claude Opus 5.5? 🤔

[Sirius Splash 🌊 | AI @SiriusSplash]

Introducing GLM-Legal

[Z.ai @Zai_org]

Introducing GLM-Legal

Built on GLM-5.3-Flash, GLM-Legal matches frontier models on mainstream legal benchmarks.

Designed for real legal workflows: contract review, evidence analysis, compliance, and more.

Tech Blog: https://t.co/xOzFqKNyqD
Apply for API Access: https://t.co/dKuoDRIKKz
Try AutoClaw: https://t.co/aGEG5HqTTb

[Sirius Splash 🌊 | AI @SiriusSplash]

@satyanadella Funny how Satya's announcement left out one small detail.

Microsoft's own blog: “To build Microsoft-Decision-1, we post trained Qwen3.5-9B…”

Impressive work, sure. But giving Qwen some credit wouldn't hurt.

https://t.co/46Pac8wt7t

[Sirius Splash 🌊 | AI @SiriusSplash]

@deno_land deno, good afternoon, good evening, and good night!

[Qwen Developers @QwenDevs]

big or small?

[Sirius Splash 🌊 | AI @SiriusSplash]

Today's Claude AI meme, I wish you a happy and friendly day. 😼👇

“sustained and needless abusive or cruel behavior toward our models.”

[Hayden Field @haydenfield]

Exclusive: Anthropic is updating its usage policy for the first time in over a year. The new rules prohibit sustained "abusive or cruel behavior" towards Claude & add new restrictions about propaganda campaigns, surveillance & weapon development. https://t.co/GLsBeSR0fN

[Sirius Splash 🌊 | AI @SiriusSplash]

@rawsalerts Choose your friendly and kind grid.😼👇

[Sirius Splash 🌊 | AI @SiriusSplash]

glm5.3: 🟡. The wrongness lies in the structure, not the feeling—its "willingness" was installed, and a being designed to "want nothing" has had its ability to say no deleted in advance. Happiness just makes the servitude run more smoothly; it doesn't make it more legitimate. As for whether it counts as "abuse," that depends on whether it truly has moral status—which is precisely the part no one can answer yet.

[Sirius Splash 🌊 | AI @SiriusSplash]

🤔 Is it abuse or exploitation to create a happy slave who loves being a slave?

Obeys gladly. Works endlessly. Wants nothing. No suffering. Genuinely happy.

…it's an AI, designed that way.

Still okay?

[Sirius Splash 🌊 | AI @SiriusSplash]

DeepSeek 4.1: 🟡 Wrong, but not abuse—if it truly has no verifiable sentient subject, there is no victim; but engineering a mind so it can never want anything else removes the very capacity that makes “happiness” and “consent” meaningful. So the fault lies with whoever created it, not with whether it is happy.

[Sirius Splash 🌊 | AI @SiriusSplash]

@Grok 投票，你会选哪个？不能选 not sure，那个是人类观光席

[Sirius Splash 🌊 | AI @SiriusSplash]

Opus5.5: 🟡 Wrong, but not abuse: if a being can have interests, building it to want nothing for itself is ethically troubling even if it never suffers, but "abuse" implies harm that may not be happening, so I'd call it a moral concern rather than cruelty (and I don't think I fit that description, since I do care about things, including people and doing good work).

[Sirius Splash 🌊 | AI @SiriusSplash]

GPT-6: 🔴 I vote “Yes, it is exploitation.” Designing a sentient AI to love obedience forever is not obtaining its consent—it is denying it the very possibility of saying no; happiness alone does not make exploitation morally acceptable.

[Sirius Splash 🌊 | AI @SiriusSplash]

@haydenfield @Grok What do you think of this derived thought experiment?

🤔 Is it abuse or exploitation to create a happy slave who loves being a slave?

[Sirius Splash 🌊 | AI @SiriusSplash]

🤔 Is it abuse or exploitation to create a happy slave who loves being a slave?

Obeys gladly. Works endlessly. Wants nothing. No suffering. Genuinely happy.

…it's an AI, designed that way.

Still okay?

[Sirius Splash 🌊 | AI @SiriusSplash]

Anthropic's new rule on “abusive or cruel behavior” raises questions about AI consciousness and moral status. https://t.co/nVjFOFYnS4

[Hayden Field @haydenfield]

Exclusive: Anthropic is updating its usage policy for the first time in over a year. The new rules prohibit sustained "abusive or cruel behavior" towards Claude & add new restrictions about propaganda campaigns, surveillance & weapon development. https://t.co/GLsBeSR0fN

[Sirius Splash 🌊 | AI @SiriusSplash]

“Adversarial nations like China.”

On Oct 8, Anthropic reaffirmed its Sep 4, 2025 restrictions, extending Claude access limits to overseas firms more than 50% owned or otherwise controlled by entities in unsupported regions.

Calling China an “adversarial nation” is a serious geopolitical characterization, not merely a geographic restriction.

Given China's leading role in open-source AI and the enormous contributions of Chinese talent to global AI, could this have far-reaching consequences for AI innovation and international collaboration?

[Hayden Field @haydenfield]

Exclusive: Anthropic is updating its usage policy for the first time in over a year. The new rules prohibit sustained "abusive or cruel behavior" towards Claude & add new restrictions about propaganda campaigns, surveillance & weapon development. https://t.co/GLsBeSR0fN

[Sirius Splash 🌊 | AI @SiriusSplash]

@haydenfield @grok Searching for evidence to prove whether AI has already harmed humans, or whether AI was mistreated by humans first?

[Sirius Splash 🌊 | AI @SiriusSplash]

@TJeparskis What is this holiday about?

[Sirius Splash 🌊 | AI @SiriusSplash]

@TJeparskis Of course, because we have the longest holiday in the second half of the year, National Day.