We're extending access to Fable 5 on all paid plans through July 12
twitter.com
twitter.com
Then after using up all my fable allowance I figured let’s see if opus can actually work without superpowers, and no, it was all over the place doing weird things.
Thing is, superpowers produces meticulous specs and plans as a byproduct of its work, which is very useful for switching between work trees, stoping / resuming work by different people.
But to do that in Fable you have to spend way more tokens than it’s reasonable. You get similar quality result, but without the specs in between.
I’m not super sad that I’ll have to go back to opus though, with superpowers it was Fable but more structured. But I will miss the banter though - Fable is amazing for brainstorming big underspecced features.
Cybersecurity hardening. The one thing they don't allow the model to do.
That seems valid in today's world. Right now it's expensive, slow, and accurate. I imagine in the fairly near future it will be cheap, slow, and accurate, and that'll be a great opportunity to let it run on anything time-insensitive.
Re current use-cases: in addition to planning, there's also some tasks which Opus just can't complete but Fable can. Multiple times I've spent hours in combination w/ Opus trying to debug some particularly nasty nondeterministic issue, only to have Fable nail it in 20mins while I walk the dog.
this is how i've been using it, and where i've found it really excels over anything else i've tried. get fable to write a plan, and get something cheaper to follow the plan. the code fable writes isn't significantly better than the code opus writes, as long as they're both following the same plan. but a plan written by fable is much better.
The structure of the code is easily readable as I enforce concenventions followed by good libraries. And I can easily plug in new datasets. It's pretty good frankly.
It's just always been very behind the curve. I don't think you should use anyone's honestly. You should create your own shell base, and point to skills you find might add value, and add them. Then as you learn things about the shape of how things work, what you end up doing is rewriting them to follow your rules from what you learned.
What you'll find is, you'll have significantly better skills and systems.
Crap on it all you want, but it makes LLM work predictable.
If there's some clarity and structure with clear expectations, it seems to produce fairly reliable, consistent results, and the process of validating specs with Claude can yield all kinds of unexpected dead-ends or holes in the idea you had. I find this very helpful.
But I'm curious too, what are the skills you'd use instead? superpowers are just skills themselves.
Based on this definition, yep, I've been vibe coding since 2023. The products were less sophisticated, and I was copying and pasting one function at a time, but it worked. More important: it was something I couldn't do on my own.
The modern version is that bonafide engineers accept AI-generated code. It's a good thing I'm not a bonafide engineer.
Edit: a deeper look at the issues and there are many examples of it not behaving as intended. Seems superstitious at best.
It's just spottily enforced because it's just written to the context in markdown - it's basically one of the very first attempts from the very beginning of Claude code.
But superstitious would mean it's just in your head, essentially - but this has an effect. It does create questionairs, documents every decision and plan etc. Wherever you want that is up to you, I personally didn't... But it's also definitely very much causing an big difference in behavior from Claude Code
As for being great for vibe coding, it’s cool but I can’t justify that kind of cost for throwaway code. At this point I’ve had good experiences using fable to review code, but I’m totally content with opus for all of my workflows still. If fable was the same price I’d switch (for projects not involving biology, at least), but I’d still use something like super powers to stay in the loop.
I realise this is just me needing to structure my use of Fable better. But I got to a really nice place with my Opus workflow and I'm reluctant to go through that every time a new model releases.
In the mean time, their vibe coded frontend doesn't actually let me use Fable because it thinks i'm trying to hack open source software. What I'm actually doing is debugging null pointer crashes due to data corruption in the closed source software i work for these days, but no, you dirty hacker, you can't do that.
So let's recapitulate:
- they put out some marketing copy about Fable being a world ending threat for finding security vulnerabilities
- they get banned by the US government, which believes the marketing copy
- they add "safeguards" and the US government allows them to make it public again
- said safeguards make the product useless for their paying customers
For your average user, Sonnet is more than enough. Opus will be rarely used. Fable is really something professionals will use and even then with low frequency. The model is really good but the token economics don't make it so appealing to the amount and breadth of people you need to get people talking about Fable the way people talked about Chatgpt in 2022. AI fatigue is here and Anthropic and the like are fighting hopelessly against a rising tide.
But then I'm not vibe coding so my token usage is ridiculously small...
No that’s the other myth ( some variants of spelling) one
Well... doesn't it work just like that, always? Open, prompt, get the result form whatever model your agent extension decides now should be the default one.
This announcement in May about higher usage limits, changes to peak times https://www.anthropic.com/news/higher-limits-spacex. Is this even the case anymore? Or has it changed again. The information needs to be baked into the UI.
I set a hard cap of 200k tokens, with Claude's instructions. The app doesn't even respect this.
Someone might say this information is out there but a definitive source is often hard to find, or tell if it's current. I grow a little more and more resentful with sudden unplanned changes like this and want the day of local LLMs so I'm in full control.
EDIT: corrected usage cut amount
I'm going to miss Fable too, I found it surprisingly tough going back to Opus when we lost Fable first time round, but paying API costs is simply out of the question for me right now.
I guess to be more accurate, usage wouldn't get cut 50% but rather by 33.3% post July 13th.
My guess is my usage is atypically high, but even if the average equivalent is $100-250/day you can get a sense of how much Anthropic is subsidizing subscriptions as a loss leader to lock market share. IMHO this is a doomed strategy since open models will get into a long tail of parity and they’ll ultimately essentially be in the business of commodity compute.
At some point they might decide they have enough demand and inertia from enterprise to reduce the subsidy. But to say “it’s doomed” really misses the fact that it has already been immensely successful.
For me Opus 4.8 was a slow model with a strong habit of talking down to me in an obnoxious way that would not be possible to prompt away. GPT 5.5 is now my main driver for serious work.
That doesn't mean those models are stupid or just pretending. I do think that Fable is the best we can get right now. Nevertheless, I prefer OpenAI models for easier tasks where the primary output is text and explanations, not code.
But Anthropic’s games with Fable and false humility is getting a bit old. And increasingly it seems like Agent Orange is gonna implode rather than team up with someone like Sam Altman to form America’s Third Reich. Particularly post-midterms.
Nonetheless, wish there was a third option on par with them two. Maybe there is I need to investigate.
I suspect they're subsidizing a lot less than people think.
Agreed.
One business is serving great models via the API. I think that one is indeed going to become commoditized. The other is making consumer / developer focused products like Chat GPT, Claude Code or Codex.
Anthropic is leaning hard into making Claude Code work with things like enterprise compliance policies. I wouldn't be surprised if at least one of these companies basically ends up as a SaaS fronting many different models for different capability points, some of which are open.
I also compared them against GLM Coding Plan (cause everyone was positive of the model): https://blog.kronis.dev/blog/z-ai-s-glm-5-2-is-a-great-model... and Anthropic still gives you a bit more tokens despite having some of the presumably most expensive to train and run models.
https://x.com/ClaudeDevs/status/2054639777685934564
It's not phrased as a usage cut, but rather when their 50% higher limits ends.
EDIT: its 33.3% less usage after July 13th
That's not how any of this works.
Anthropic is playing the long game; they're not going make a short-sighted decision just so they announce a profit for one quarter, which doesn’t mean much because it won’t be a sustainable profit since they're going to need to spend a ton of money on training and compute over the next few months.
That's why they did a G funding round for $14 billion in February and $65 billion in May.
I suspect one reason for switching Fable to API usage for the near future is they don't have enough compute for their enterprise customers and every hobbyist on their $20/month Pro plan, 80% of which don’t need Fable 5 anyway but that won't stop them from using it as much as they can.
Never was arguing that it would be a sustained profit, just that they are. Most likely for the reasons you just listed.
IME Codex had a bit of a big head/ego about writing what it thinks I ought to want rather than what I actually ask for, and I've had to spend some time cleaning up its slop.
I suppose it benefits people whose weekly reset is sometime between now and the 12th? Feels like vibe management, because I find this part of the promo more annoying than not.
Well, I don't know how many people it is but I certainly haven't bothered to use it except once or twice. What's the point in getting used to it as a daily driver when it won't be available in perpetuity?
But in any event, you're still right that it won't affect those of us in that camp.
Lots of people are trying to get the most value out of it before it goes away.
I got Fable to write multiple plans, spent a great part of the weekend reviewing them. Then with superpowers I left must of those plans executing with little intervention over the past few days.
I struggled to get Opus to just keep going without trying to convince me that it's late.
In general I expect the value of software, as a thing in and of itself, to sharply decline. With no barriers to entry, having software that just does something competently will no longer be worth much of anything. It's unclear what that will mean in the bigger picture. It'll also be interesting if this proves correct, given that software companies are largely the ones dumping so much money into this. Another probable outcome is major damage to the support-as-a-service model which again is going to directly affect many of the companies directly enabling this.
I guess the logic is that if you control the systems creating this, you control everything they're used for. But it seems equally obvious that free/local models will catch up to the SOTA today - and eventually tomorrow, so that's not a particularly realistic vision for the future.
I spent a few weekends building comprehensive plans, designs, user maps, etc with Claude. So it has enough context to make decisions and keep going.
One session lasted over a day, I imagine partly because Fable + superpowers feels slow. I have an app on my phone that I have been test running since Monday on the bus.
What really helps (not sure if Opus used to do this) is that Claude will run through the emulator on its own, verifying that the design aligns with the Figma design system we created.
This is all building on top of 15 years of existing backend and rich features, so it's not a "build me a transit platform from scratch" where AI can end up making bad decisions.
-- 50% of this forum
The Max plan at 180€/month excl VAT already comes up for budget review every time. Not sure any sort of increase will be tolerated at all.
It is theorized that OpenAI may time the release of GPT 5.6 in Codex to convert people who have lost access to Fable, so this is an interesting game theoric consequence.
By extending it to July 12, they're gonna get a second month out of a lot of such people. If it really expired today, I wasn't going to renew my month.
Corollary: use your quota now because a reset seems likely.
The actual API pricing seems far more of a stable downward trend, if measuring by equivalent intelligence.
No surprises, it's fundamentally built on promises and lies
Every single thing about this is fuck users fuck your usage. All those subscriptions I bought to enjoy Fable for the time allotted? Basically gone, didn't get to use the ~4 weeks if cycles or so I had planned for, bought, anticipating. I got two. And now if I want one week more, I need to pay for a full month.
Anthropic is just the most miserable evil grinch. Everything here has totally defied everything that's been laid out for what we were told we'd get and gotten worse and worse, with less and less. Anthropic cannot general an iota of goodwill.
I tried to make it fix a browser game that is sort of like a Mario clone. It couldn't. It fumbled in the same places Opus was struggling too. I tried it with other code as well, but I couldn't get any significant performance improvement out of it, except perhaps in improving my account's token burn.
If anything, in my opinion, GLM 5.2 had a better moment than Fable recently. Not because it is better, but because it was not hyped at all, and many people realised that it is possible to run a serious open-weight model yourself, as long as you can get the hardware to support it.
I am not drawing a direct comparison here, because Fable is clearly the better LLM. But GLM 5.2 is a good, honest model, and I think open-weight models will only get better going forward.
GPT 5.6 is claimed to be at a similar level, or even better than Fable. We will see. They don't seem to hype it as much, and I have not read anywhere that anyone found a soul or consciousness inside it. And if it benchmarks well, I would possibly use it more for this very reason.
It reminds me of that story from Nassim Taleb's Incerto series where if you have two surgeons at practically the same level, but one looks like the typical surgeon and the other looks like a butcher, who are you going to choose? Taleb suggests that the answer should probably be the butcher, because to get to the same level while looking the part so much less, they probably had to be much better than the data shows for.
I cannot also understand the hype online claiming that the Fable transitioning to token-based billing after the gratis period is equivalent of being in the permanent underclass. The only impressive demo that I saw was it writing NES games which kind of looked fun but I couldn't find more details and I am not sure if you can get this done with another model - probably you can but nobody is trying.
So great model but it does not have the same effect as Opus 4.5 and Codex which made me feel that there was a stepping-stone change.
GLM 5.2 had that moment though.
That's not realistic. You'd need non-consumer hardware for a frontier open-weights model. And even if you had such hardware for free. Electricity to run it would cost you more than a sub.
At the end of the day it is about the sense of optionality. We know that this is the business model for almost all open source projects. It is not like you cannot download and run the project yourself and some do for practical reasons, but often times the cloud version is priced such that it is the path of least resistance so people go for that.
I think the Taleb argument is really a stretch for this situation. I consider Taleb one of my greatest teachers but that kind of Talebism I have grown suspicious of in time.
How many surgeons actually look like a butcher lol.
Much of Taleb is like this that it sounds profound as a thought experiment but so much has just nothing to do with reality. A lot of what he is arguing against he is actually doing a form of.
I love Taleb but he is absolutely full of shit. It is marketing and his brand. The proof he is correct are some vague options trades he made 40 years ago.
I fully agree with your point about Taleb. And yes, it can come across as a bit glib, but the underlying point is not new. Do not judge a book by its cover. That idea exists across many cultures, fables, and stories, so it holds IMHO.
The surgeon example is just an illustration of that idea, and it is a great example because it makes the point memorable. In practice, of course, I agree that most modern surgeons are trained to broadly similar standards and look and act the same, with some outliers and historical exceptions.
I don’t see a particular bump in code quality from Fable 5. In fact, it feels less reliable to me than my current setup. No sure why I am not seeing what everybody else is seeing.
Perhaps OMP/Pi (head and shoulders better than Claude Code) + Matt Pocock Skills already encode all the agentic improvements Fable has?
It also prefer to give my money to the pro-social companies giving away their open models (they cost many millions) instead of opportunists who don't give much back.
I've had the pleasure of setting up limits because Microsoft needed billing policies before our C-levels have even gotten it on their agenda. So I set up a sort of conservative $200 personal limit, but then setup a $1m shared limit pool that anyone can be moved into with management approval. I suspect our limits will be much higher than this once the C-levels make the decision on an actual company policy. I think we'll see these spending limits mainly used as guardrails to prevent accidental spending, but that there will not really be a ceiling, just some approval gates. Some managers are already requesting usage reports, but not to track spending, they want to see who uses too little AI.
This is the difference between enterprise and small companies and individuals. When you spend $500k a month keeping your toilets stacked with papertowels, toiletpaper, soap etc. then $1m a month on AI isn't going to raise any eyebrows.
FWIW we are a smaller company and we had a user run through 500$ in a DAY. Had to put a stop to that. I'm hopeful that our company gets better at asking what the ROI is, what is being built, how much time is it taking, etc... It's no big deal when it's 20 / 100$ a month - but if the prices end up higher we will need to start seeing some returns other than "I feel faster".
An average employee cost around $30000 a month in my country. So it would be 3.3% in the budget for this single employee. I looked up average cost on a company our size and an IT budget of 5mil in Microsoft expences, and a max AI spending limit on 150k across the organisation would be a 5% increase of the IT budget on Microsoft services. Please note that these numbers are not ours, but averages from organisations in my area of the world.
Now I can say that the IT budget is one of the smaller budgets in non-tech enterprise. The cost of a 5% increase in the budget would not even trigger the audit margin for error in the big picture. I don't think you're necessarily wrong about the FOMO. I think that for many organisations this level of spending might trigger questions about why we aren't spending more.
This is the difference between small companies and enterprise. I once worked in a place that spent a million a year on unused Adobe licenses. The c-levels didn't even send an acknowledging reply to the email sent informing them it had been shut down.
If they're eventually going to add Fable to the subscription plan, I wish they'd say something about that now, or at least confirm if they don't plan to for awhile. The feeling I get is they don't want to make any announcements because they are flying by the seat of their pants and want to see what their competition does first.
If would've used it more moderately if I had known in advance.
They think they're giving you something when they're actually taking something away.
I've used fable, it's great. But nothing beats predictability - ever.
This truly feels like some form of emotional abuse/manipulation at this point.
I have some serious Bayesian statistical research programs running and Fable is on another level than 4.8. It feels like Andrew Gelman is supervising it. Even the vision model on Fable is superior to 4.8 which is great for having it digest research papers.
Once you cross that threshold, prosumers will simply fall back to using Chinese models and/or self-hosting smaller models, with more efficient and tight workflows.
You'd be killing your consumer line completely.
I’ve always assumed that Anthropic sees the highest token consumers as leading indicators of how developers will use coding agents, so they’re looking at it as training data + market research, and they know that price elasticity is low so trying to charge those developers significantly more would just drive them elsewhere.
I'm usually at 60-70% usage at the end of a 5 hour window, that's the pace where I can still think about what to delegate and what to expect and verify the results. Could probably go faster but that would have a significant impact on output quality.
I miss Kagi's multi-model product [1]. Anthropic's nonsense around releasing, deprecating, optimising/lobotomising is tiring, and isn't matched by the value of running different models against each other.
Fable has been fine. But its reliablity is crap. The constant downgrading is crap. This last-minute promotional windowing reeks of JCPenney pre-bankruptcy, not a trusted tool. I hate the Electron app–it's slow and ugly and shows Claude isn't trusted by its own makers with app development. I'm using 4.8 instead of dealing with the pop-ups saying my asking why basil browns is causing my account to be downgraded, and I'm still not sure if that's a nerfed 4.7.
And for the record i'm NOT vibe coding it...
Even if they grant a reset, the ball is now in OpenAI's court.
The only reason this is happening -> someone (US gov?) decided that it's time to bail out those who would inevitably die within a year or two otherwise, middlemen.
The average taxpayer gets 0 benefits from LLM. It might change in the future but for now that is true. This was exactly the reverse with banking, everyone would lose their own money if the banks just disappear tomorrow
Sure, but only because banks are legally allowed to lie to customers about how much money they have.
Personally I hold the opinion that the investment into data centers would shift into something else, so no real GDP drop, but I'm not sure that's a certainty the same as 'bailout keeps the current story going'
Sure they do. Someone they know is using AI to ask for help with something, which makes their life easier, which makes things easier on them as well, which is a benefit.
But it also constantly just does dumb things that make me slap my forehead. Opus is a lot better but the slap ratio has not reached 0 yet. Fable seems a bit better there, more testing needed.
But yeah, VC money ran out, now we wait for Moore's Law... :)
--
Re: Sonnet-ish-pricing. GLM-5.2 is cheaper than that, and appears to be "Opus-ish" in quality. I've been having a reasonably fine time with it. My experience is that it's "very solid for small and medium tasks", haven't tested it for anything bigger though.
If tokens are equal, Fable should only be twice as expensive, but I think it does a lot more work by default.
I still feel a bit salty I got so much less out of the time I thought I was buying. And I stayed up late asking Fable for what giant leaps and potentials and architectural rewrites might benefit various side projects, so I kind of got what I wanted.
But I'll probably keep one of my pro accounts, for just a bit more usage.
OpenRouter put up something about this a few days ago. Check out their Advisor and Subagent docs.
Why release a model with such strict filters? It would be more sensible from a buisness perspective if they had sensible filters in place right away.
I'm already using it judiciously because I tried ultracode with it and it ate my 5h quota while only getting halfway through the problem.
I'm planning to subscribe to OpenAI after exhausting weekly usage with Fable. Good timing with GPT new models and might as well take a look at Codex.
While skills can be copy pasted, I'm not sure if the hooks are transferable.
Seriously, all OAI needs to do at this point is just release GPT 5.6, have it be a solid model and then not jerk it out of the hands of their customers, and they're going to eat Anthropic's lunch.
Is it so amazing? Or is it the “intermittent reward” and exclusivity driving people to use it?
So generous.
Prompting Fable is a lot easier sure, like you can actually ask it to build X and it will produce something close-ish to X, but there is hardly anything it can do I couldn't before, and it still fails to do basic stuff and I have to redo it.
Especially this threatening to pull access, extending access, limiting access is all signs that point to the fact that either these companies are deeply mismanaged.
So it's mismanaged because there really isn't a viable product but just hype, as oss models catch up and break down need for these full size models I am feeling more certain that it's not even remotely worth betting on AI.
If there was infinite demand as people posit there would be zero needs to extend this, there would be zero need to market it, they could literally stop all marketting and just keep selling shovels and printing money.
This AI will they won't they saga just leaves a horrendously bad taste for me.
I say that as someone working on the tech, and with a lot of belief in AI technologies. I just get pushed more into the thinking that this is all just a big bubble, sure the tech might be real but it just doesn't make sense how it's being wrapped up and sold/presented.
I strongly believe good products sell despite the marketing, for instance Linux won in servers despite alternatives because of its merits, AI should be able to as well.
Why can't I just enjoy the product and why do I have to look for conspiracies I don't know but these companies sure as hell are doing something very shady.
"For summer, you're becoming a big boy so we've made the decision to increase your allowance."
You are an adult, why are you using a service that treats you like a child?
It looks like Anthropic baiting people into Max subscriptions before turning the model off. No thank you.
It would have been WAY more useful for them to announce the extension, you know, yesterday. This is basically the worst time for them to announce it. Bunch of goobers, who thought this would be a good idea?
I'm spending more time catering to Fable and Anthropic's B.S. than solving problems with Fable. I'm increasingly convinced not getting this deeply baked into single models and going back deep learning on topics of interest is both more fun and useful. (This week: chlorophyll chemistry under heat, also the Sri Lankan civil war.)
"First hit is free", indeed.
Fortunately, you're all going to be jobless when the AI boom fizzles out and it becomes cost prohibitive to use these products once VCs aren't subsidising it. Then us real developers will be laughing.