A few words on AI:
1. First of all, it isn't 'AI'. It's (just) simulated intelligence. Which is what it will tell you if you will ask. "The beginning of knowledge is to call all things by their right names."
2. LLM's are about as good as they ever are going to be.
3. Anthropic lost sight of the goal. They tapped the brake in the middle of the race when they should have been jamming on the accelerator. It may cost them the entire market and they just get acquired.
Imaging Henry Ford saying, "We need to spend massive resources making sure the vehicle doesn't go faster than a horse, for safety's sake". - Someone else leaves him behind.
A. Not only is it NOT true that they have the best "security" model (Microsoft cobbled together many smaller models and achieved the same result. Numerous 'smaller' models have shown identical results. It's all about what you point the model at and for how long.
B. Fable is only much more compute, for longer. It isn't clear that it is any better than Opus 4.5.
C. There has been no "Wow!" AI launch since Opus 4.5 in Nov 2025. Hence: they have hit a wall with the current paradigm. The system is no closer to "Jarvis" or "Samantha" than it was in Nov 2025. Any improvements have been improvements in harness and/or 'thinking' (simulated thinking) time.
4. Anthropic Pt 2: All the marketing bs about "it's alive!" and "It's so dangerous!" and "The govt must step in!" (after which, the govt did) as well as horrible customer communication gave up the market to OAI. And there is zero loyalty to an AI brand. People switch with one click
5. Even more humorous than that: Now GLM 5.2, Kimi 3, Hy3, MiniMax M3, Deepseek 4, and a few more (I can barely keep up) are either rapidly closing in on SOTA of US flagships or even surpassing them in some ways.
Why? Because the tech growth is asymptotic. We are all at that line now. There is no 'much smarter/better" version available because the tech has hit its limit.
Now, it's a matter of opinion, style, refusing to refuse, "Imagine if your hammer turned into a tiny screwdriver while you were swinging it. gfy OAI and Anthro--many dangerous things are sold every day. You don't need to childproof it MORE THAN YOU HAVE." -- Fable was unusable for me. OAI is switching me to Luna low from Sol Hi with no way to stop it.
THOSE ARE ENOUGH TO MAKE ME LEAVE YOU BOTH PERMANENTLY. Why?
I can't trust you.
The system is already far from what I need and if you silently kneecap my usage mid-run, I can't count on you at all.
Also: local models are now 80-85% as good. That run on my single GPU. Explain why I need y'all again?
PS
And it's limit, no matter the model, is still not good enough
How else do we know Anthropic et al are lying?
See the articles, "Soon, we won't need developers anymore," while they not only hire developers, but buy entire companies full of them (why don't you just vibe bun in a weekend, Dario?). By their fruits ye shall know them. Watch what they do, not what they say.
The funny thing is that if there had been any adults in the room (like there appear to be outside the US), this thing has great capacity for good.