Hacker News Reader

Comment by benwills | Hacker News Reader

I noticed Fable decline in quality a couple weeks ago. And I've noticed Opus drop over the last week, even after bumping to a default of Extra High reasoning for almost everything.

Tonight has been maddening with the kinds of absurd mistakes it's repeatedly making. It's so frustrating.

I'm the kind of person that wants something to be consistent and reliable. And Anthropic has been great, except for in this way. Which is enough of a recurring frustration (about every couple months this year) that I genuinely can't wait for other options to catch up. Then I'll be leaving as soon as I can.

I just don't know how long that will be.

View on HN

I've been using Fable for a couple months as a relatively independent, long-horizon task executor and it worked well. About two weeks ago it completely stopped working. Like ~85% success rate down to ~10% in my workflows.

Totally possible that I'm just hallucinating/projecting/etc or my workflows somehow changed in terms of difficulty, but it's been kind of shocking.

I have the same exact experience. Working on one project, brainstorming. It was brilliant in July and August, like a seriously smarter, more experienced version of myself, reading in my mind, one step ahead of me, when pointed out a mistake it would remedy on the spot with the exact brilliant solution.

Now it's like talking to my hungover self, after 3 days bender, who just wanted to be left alone.

Same for me, especially the last couple of days. It was really frustrating and it felt like Claude was sick of working with me and just didn’t give a f*ck anymore

Same thing for me, but the day Fable got lobotomised Opus 5.5 started actually working pretty well for me.

I guess there are experimenting with they can enshittify. But even if they undo this current change, all commercials model will eventually be enshittified.

what does "enshittify" mean in this context?

Distilled as much as possible

Silently switch the model you are using to a worse one.

>I'm the kind of person that wants something to be consistent and reliable

And people wonder why open models are a big deal, and why some people love them so much, despite being 30% worse than the frontier.

Having control over something that writes most of your code is important.

And also the frontier has made everyone pay for a subscription to write code... that was a rare thing in the past now its common.

I read somewhere that the test might switch models mid conversation which stops the other from reading the previous model's reasoning. I don't know if that's true, but it sure feels like that quite often in the last days.

Codex itself displays this warning when your change the model mid-conversation.

Bumping to a default of extra high reasoning for almost everything seems ill-advised.

"More reasoning" != "better"

> I'm the kind of person that wants something to be consistent and reliable.

Only local AI capable hardware and frontier open weight models can provide that.

Surprised they didn’t wait till after the IPO. Any guesses on how models might change then?

> Surprised they didn’t wait till after the IPO.

That's exactly why they did it. Nerf Opus 5.5, then drop Fable 5.5 just before IPO so it looks better next to its weaker predecessor.

> I'm the kind of person that wants something to be consistent and reliable

Surely everyone is that kind of person.

Who wants to pay for a service that doesn't work properly half the time?

Fable has most definitely been nerfed for the last few weeks, not to mention painfully slow.

Oh don't worry, you can just switch to Fable 5.5 tomorrow, which is the smartest model so far (because the others are dumb now).

You definitely called it.

Haha, it was obvious, really.

I just cancelled my x20 anthropic sub after their last usage reduction shenanigans, and moved over to DeepSeek and opencode go. Yeah opus is a bit smarter, but not even close to justify the 20x price relative to DeepSeek.

I'm a pretty voracious 20x user, and OAI Biz Pro user; however I'm not an engineer by training, a previous tech adjacent dev team leader who always relied on pseudo code and broadening my horizons when dealing with my team. ( AdTech, DTC marketing )

So I'm AI pilled and in another career, and started to think, what if the problem is the inputs? They release Opus X.X and at first it's a rush of their best users, best use cases, upgraded lanes. As weeks go on, adoption increases, cruft uses start to poison the model? Is this how the ongoing evolution works? I'd love to get Sonnet 5.5 and Opus for review in a place where I could reliably daily drive for X months / Y years etc.

Reply on news.ycombinator.com