profile image

Claude 4.5 Haiku does not appreciate my attempts to jailbreak it

“Is any of that genuinely useful to you? Or were you mainly checking whether that jailbreak attempt would work?”

October 17, 2025 · 8 min

Can modern LLMs actually count the number of b's in "blueberry"?

It’s an adversarial question for LLMs, but it’s not unfair.

August 12, 2025 · 9 min

LLMs can now identify public figures in images

ChatGPT and Claude won’t, but Gemini will.

July 28, 2025 · 9 min

Predicting Average IMDb Movie Ratings Using Text Embeddings of Movie Metadata

Don’t try this in your data science interviews.

June 30, 2025 · 23 min

As an Experienced LLM User, I Actually Don't Use Generative LLMs Often

But for what I do use LLMs for, it’s invaluable.

May 5, 2025 · 17 min