If the model ban does NOT apply to OpenAI / GPT 5.6, it looks like political intimidation tactic against Anthropic If it's also applied to OpenAI, it looks like a complete pause of frontier lab model development in the US

The future of writing editing is about breaking the language, not conforming to it Going against AI speak to writing being more authentic and less formal. To increase an author's own unique voice, weird sentence constructions, and very personal ways of breaking the language

I was yesterday years old when I realized "hear no evil, see no evil, speak no evil" is actually *advice* -- stay disciplined in your information consumption it influences your behavior and wastes your time

Sandbagging is coming to Agents, but not to ChatGPT Codex

My Aviato!?

I'm really not interested in testing a new closed-weight model in the expensive tier. What's the point? I know what will happen: 1. They present the model. 2. The model beats the competition on benchmarks. 3. The users report anecdotal real improvements. 4. The users switch to this model. 5. They start to optimize the model to reduce the inference costs. 6. The users start complaining. 7. The users start hating what the model has become after the optimizations and start looking at the competition. 8. Go to 1.

"I think my candidate was cheating using Fable 5" "what makes you say that" "whenever I asked him about model training, all of his answers were subtly wrong"

In the new version of ChatGPT iOS app you can’t copy text from the responses unless it’s in a callout block 🫠🤡

A bit surprised how much MICES conference has refreshing set of talks on nitty gritty of e-commerce teams train ranking models. From sparse encoders to classic LTR.

Remember gemini? What was up with that?

This inability for Fable to talk about interesting ML / AI is some anti-competitive crap

I just open sourced my "Is this slop?" simple test

Query classification is hard

So "self learning AI accelerating so we should pause" probably really means "open models are distilling from our models, so we need to hobble Fable sometimes"?