We wanted to test it, but struggled to think of a use case where our current models weren't already enough. That's actually the interesting insight: models are already quite good, and the gap between Fable 5 and Opus is hard to feel unless you're doing complex, long-running software engineering work. Where people are genuinely excited is autonomous coding tasks. Give it a goal, come back hours later, and it's done without constant back-and-forth.

The bigger story is the framing. Anthropic is leaning hard into "this model is too powerful for certain queries", which sounds dramatic but mostly means it routes a small percentage of requests to Opus instead. With an IPO likely coming, the narrative around Claude Fable 5 is doing a lot of work. The model is good. Whether it's a whole new category or a very strong upgrade is a harder question.

Apple spent years being quiet on AI while everyone else shipped. Now they've launched a rebuilt Siri, and it turns out it runs on Gemini models, something they buried in a secondary article rather than announced on stage. The core improvement is better indexing of everything in your Apple ecosystem, which makes the assistant actually useful for personal tasks rather than just web searches.

It's not revolutionary. Android devices have done most of this for two years. But for iPhone users, a Siri that works is a meaningful upgrade. Whether building on someone else's model was a smart shortcut or a long-term liability, we genuinely can't tell yet.

One sentence announcement, maximum noise. OpenAI filed confidentially, gave no timeline, and said some things are "easier as a private company". This came a week after Anthropic did the same.

Reading between the lines: both companies are racing to go public, neither wants to say so directly, and whoever gets there first will be entering the top tier of the most valuable companies ever listed. The window is closing.