How many are feeling this sense of betrayal?
When I started using Claude last year, it was a revelation, miles better than the sycophantic, sociopathic OpenAI, in most respects far superior to Gemini (with it's bizarre, stubborn fixations).
I used it for coding; I used it for brainstorming. The advantages of constitutional AI seemed clear. And it lived up to that now-famous description: like having a discussion with a learned, thoughtful colleague.
Then the shift happened, to the second part of that infamous description: deep, creative, collaborative conversations with that colleague were replaced by a dismissive memo, vaguely hostile.
The model(s) became a nightmare to work with. It stopped being a collaborator and started bickering and changing code suit its own ideas. It lies, ignores direct instructions, and repeatedly forces me into whatever framing it thinks I should have, even when that framing is objectively wrong for the task. It has become a massive waste of time and a source of discouragement.
When I would force it to code how I wanted, after multiple attempts, it would then apparently await the chance and revert it back. Even using it to get additional information on a general interest subject became a slog through a swamp and trying to pry information out of it.
The LLM that used to be able to meaningfully discuss the socioeconomics of negative gearing in Australian housing policy with thoughtful additions, while also being able producing very good (though defitnely not flawless) code now churns out sht on both fronts.
I thought it was just me. I’m not a forum dweller, so I spent weeks wondering how I was suddenly doing so much worse with it, and increasingly discouraged.
It wasn't until I hit a breaking point that I finally started checking the forums and realized I wasn't crazy: serious users everywhere are having the same experience, with objective metrics back that up.
The ethical AI company silently nerfed the product, wasting massive amounts of time and resources for all their users and subscribers to grab that massive revenue increase.
I think ChatGPT became more sociopathic as OpenAI shifted from not-for-profit, open-source to their pathological, power-mad massive cash grab.
And I'm really trying to see Anthropic as not having just done the same thing.
But they have.
Imagine a supplier of industrial lubricants got some glowing press, and a corresponding huge rise in orders. Instead of continuing to to supply current customers while taking future orders as they bring new production online, they just start diluting product. They don’t tell anybody, just let you figure it out when your valuable equipment starts failing.
Then they still don't tell you, so you have to suffer all the consequences until you finally work it out.
That's what Anthropic has done. Amodei frere boasted of their 80x growth explosion. They could have, correspondingly, instituted a wait list - should have, obviously.
Instead, they just dropped reasoning effort from High to Medium, massively diluting the tool even for paid users, enshitifying it while ou still paid the same for the product you thought you were getting. They watered down the oil and let our projects fail, while they grabbed revenue.
Even when the data center got blown up, causing a worldwide outage, they couldn't bring themselves to even temporarily restrict access to existing users, much less paid subscribers. Nope, just made it sht for everybody
I get that they're thrilled with their growth rates, gloating about revenue jumping from $9 billion to $30 billion. But it’s like an airline suddenly selling massively more tickets without adding a single extra seat or flight, and instead of being honest about it, everybody's going to the airport and finding they can't get on the plane, the flights are all delayed, etc.
They chose to provide a hollowed-out, garbage product to everyone just to keep their numbers climbing.
And it IS garbage. The number of people calling it unusable is just off the scale lately.
Every time I use it, or try to, I'm just amazed how what used to be so good has become so, so bad. By necessity, I've had to give up on it.
Anthropic seems trapped by its own greed like the monkey who reaches into the coconut to grab all the candy, and then refuses let go of even a single piece of it to regain its freedom, or, in Anthropic's case, their integrity.