I asked Fable 5 (Anthropic's new AI model) to make me a launch video for my Social Media site Glowwww. I didn't write a script. I didn't edit a single clip. I just told it what Glowwww is — and let it work. An hour later, this was done. Watch till the end. 👇

GPT-5.6 could be released at any time. When it does, the internet will be flooded with fake screenshots, benchmarks and release notes, and people claiming to have known first. I’ll be tracking the real updates, sources, features, benchmarks and launch details here. Follow @fanofaliens and turn on notifications.

I asked Fable 5 (Anthropic's new AI model) to make me a launch video for my Social Media site Glowwww. I didn't write a script. I didn't edit a single clip. I just told it what Glowwww is — and let it work. An hour later, this was done. Watch till the end. 👇

Claude Fable got 60% cheaper because someone took DeepSeek’s “turn text into pixels” idea and used it to make Claude read code screenshots.

OpenAI seems to be quietly testing GPT-5.6 in Codex. Beta users can now select Sol, Terra, and Luna from the model picker, though most users still can’t access them. Real-time voice functionality is still under development.

Just a heads up: Fable 5 access through Claude subscriptions will end on July 7. This is not a shutdown. Anthropic says demand is currently higher than the available compute. For now, Fable 5 will move to pay-per-use usage credits. Once capacity expands, Anthropic plans to bring it back to subscriptions.

Anthropic relaunched Fable 5 after nearly three weeks, but safety filters are already pushing a small portion of requests to Opus 4.8. People waited for Fable 5 because they wanted Fable 5, not a weaker substitute chosen for them without warning. Also after 1 week they’re going to switch to usage based credits for the same performance. I really want OpenAI to win here by launching all GPT 5.6 models asap to counter anthropic.

just the bat and his cat

Everyone is talking about Claude Fable 5, but this is a really important breakthrough that is being overlooked. OpenAI reportedly found a way to make ChatGPT much cheaper to run, with some responses costing more than 50% less to generate. That means OpenAI can serve more people using the same compute.