FYI, I casually tried to reproduce this on three writing samples from many years ago, which I do not believe to be on the public Internet and which were varying subjective difficulty levels. Opus 4.7 went 1 for 3 with ~no effort in the prompt (~two sentences long).
LLMS
-
Meta’s JEPA Strategy Shift: LLM Focus Over World Models
By
–
Right. Except that Mark Z, @boztank and others in the leadership were always supportive of the JEPA / World Models project as a long-term bet.
But the AI strategy of company became more LLM-pilled and short-term focused.
And many of the applications of JEPA/WM are in industrial -
ColBERT Memes Trend Takes Over Social Media Timeline
By
–
My timeline is full of ColBERT memes. How did that happen?
-
New class added to lisette LLM library for easy integration
By
–
Thank you! 🙂 I have added a class to our lisette LLM library now to make this easy to use too.
-
Liquid Hiring at ICLR2026: Join Our AI Team
By
–
I'll be in Rio for #ICLR2026. Let's meet at the Liquid booth if you're around! We're also hiring for many positions across applied post-training, inference, pre-training, infrastructure, multi-modal, and many more!
-
Evaluating Local AI Models on Mac Mini Hardware
By
–
Got a Mac Mini ready that's running Hermes. Truth be told, I don't know how to evaluate these things anymore and don't have many hours to put into them. How should I judge it via OpenClaw, Hermes, NanoClaw, Zo, and others?
-
Google AI Subscription Tiers and Usage Limits Explained
By
–
On 1, the simple mental model is low, higher, highest wrt limits for Google AI subs, we don't publish the specific numbers because it is all changing and shifting over time. But Ultra users get the highest limits we offer, Pro is less, and free is even lower or zero. On 2, this
-
Qwen3.5-Omni: Fully Omnimodal AI Model with Thinker-Talker Architecture
By
–
“Qwen3.5-Omni Technical Report” Qwen3.5-Omni is a fully omnimodal model that understands text, images, video, and audio, then responds with text or real-time speech. The architecture they use is a Thinker-Talker split, where the Thinker performs multimodal reasoning, while the
-
OpenAI announces ChatGPT Images 2.0 with improved accuracy and text rendering
By
–
ChatGPT Images 2.0 is a step change in detailed instruction following, placing and relating objects accurately, and rendering dense text, with the ability to generate across aspect ratios. It’s also accurate across languages and uses its expanded visual and world knowledge to