This is very surprising. I've never seen a legitimate business not give refunds for technical errors of their own fault. Minimum Anthropic should credit the full amount to them.
"You're totally right! I'm sorry but you're going to have to piss off anyway. Would you like to spend a few more hours discussing it with our AI chatbot? It won't help. But if it makes you feel better, it will probably cost us an extra $0.12 in tokens."
I'll bet the first human at Anthropic learns about this from HN.
I’m coming up on my one year anniversary of having my Claude Pro account terminated for reasons that to this day remain an utter mystery. “Here, submit this Google form and we’ll look at it.” They have never done so in the one year since this happened. Once I interacted with what seemed like a human; but weeks later it was replaced with the brain dead fin.ai
At least they did not steal my money; so I should be grateful for that. But as a small potatoes user, I advise everyone contemplating dealing with this user-disrespecting company to walk away.
This issue would have never gotten a response if it didn't go viral.
My StubHub story: bought $500 tickets and accidentally bought ones in the dsabled seating section. Called 2 minutes after purchase when I realized - their response "you can relist them on the site". Who else was going to buy them?? Nobody did. Any normal human business would let you correct a basic human mistake like this, not even 10 mins after purchase, but not stubhub. They could have upsold me and I probably would have left happy! At least I could have attended. Cost me $500 but cost them a lifetime of emnity
Of course, I suspect the true business model to be to do nothing. You sell the "service" to people customers, but your enterprise customers pay you a subscription fee to not execute the order. ELaaS: Everybody Loses as a Service
https://arstechnica.com/tech-policy/2024/02/air-canada-must-...
Small claims court is exempt from arbitration requirements (which are primarily aimed at avoiding class action suits). It doesn't require you to hire a lawyer, and probably won't get your account automatically nuked the way a credit-card chargeback would.
So, you can waste as much of their money as they wasted of yours.
This is Anthropics initial response, which they walked back ONLY because of the HN outrage. Without HN, that would've been tge official answer.
I'll judge them on that, thank you.
May be somebody will start a business where such high-value-per-employee companies could outsource customer support to be performed by real humans? ... And then such business would replace the employees with AI agents ... It is a trap.
My guess is this response was entirely written by an LLM that is instructed to never to offer refunds or compensation.
They are trying to go public and will get absolutely bitchslapped by SOX.
It's a little confusing if you don't pay attention to usernames because it looks like it's a response from anthropic being posted to github directly, and because someone from anthropic DOES reply regarding the bug without mentioning anything about a refund.
Claude Code support admits that Anthropic has a policy to defraud customers
Dude what is it with HN and using extra soft words that don't at all mean the actual thing they're supposed to mean.
Nothing there is a surprise.
This is very bullshit and probably (in a better world for sure) very illegal. Can't bill more than you've actually delivered and what the customer in advance agreed on.
Stop with this god-awful corporate-washed lingo. You're not being professional, you're skewing reality.
So giving this guy his 200$ back would open the flood gates for other such requests. Their behaviour, as much as it is weird and antisocial, perhaps even breaking some laws - is completely logical in their own weird world of "Datacenter PhD nation" or whatever other bullshit they use to hype up their "product".
I investigated. I was being for a Claude Max gift subscription that has been sent to – what appears to be – a randomly generated 27 char alphanumeric icloud email account that bounces.
Apparently, Anthropic doesn't have a centralized process that allows you to approve, see or revoke "gift cards." And no I can't use this hypothetical gift card. Because I can't see what the system generated, when it generated it, and if the "gift" sent to this 27 character alphanumeric string was redeemed.
Their support bot doesn't work. As it's a possibly suspicious charge (I certainly didn't buy it), I've been trying to get them to revoke it. But the bot passes it to a human and their humans just close the ticket without comment.
I realize that people working at Anthropic are "just" researchers building cutting edge models. And that Claude is really great and all. But hasn't anyone told them about the global legal risk of incorrectly billing millions of people?
What is their legal risk team doing? Their ops team? Or, whoever else is responsible. Even their own models, Opus 4.6, Opus 4.5 and so on will flag this as a legal risk on "max" thinking.
Because even if $200M to $20M seems "insignificant" next to the however many billions they made in the quarter. Knowingly perpetuating fraudulent billing practises is a real legal risk with real prosecutorial (and financial) consequences. It's absurd to me that so much of legal risk analysis fixates on how users use the tools they pay for, but not what's an obvious trigger for class action lawsuits and prosecutorial investigations in most jurisdictions.
This isn't even a threat. The FTC has taken Uber to court, https://www.ftc.gov/legal-library/browse/cases-proceedings/2... and is apparently seeking a few billion in fines?
https://www.independent.co.uk/us/money/uber-lawsuit-fines-bi...
Purposeful unauthorized billing was found to be fraudulent and defendants were made to fork over assets, https://www.ftc.gov/news-events/news/press-releases/2024/09/...
And this is government action in one jurisdiction. The EU has fines of 6% of global turnover, and yes, they too will seize assets if the fines are unpaid.
What I'm trying to say politely is, does the Anthropic team realize this is an insane legal risk. And to quote Trevor Moore's immortal words, "insanely illegal."
Why would you do this? Does anyone realize the implications of this? At all? Other than the AI models that the humans aren't paying attention to?
screenshots for anyone interested, https://x.com/_areoform/status/2048644232043434354
I contacted my bank and got a reply (from a human) that all three payments are valid.
Emails from Anthropic state that the first two payments failed, but the third went through. Fin says that my question will be elevated to a human being, but so far I was not contacted.
* notices@anthropic.com : from https://www.anthropic.com/legal/commercial-terms)
* usersafety@anthropic.com : from https://www.anthropic.com/legal/aup
* marketing@anthropic.com : https://www.anthropic.com/legal/consumer-terms
* disclosure@anthropic.com : from https://www.anthropic.com/responsible-disclosure-policy
* dpo@anthropic.com : from https://www.anthropic.com/legal/privacy
* pubsec@anthropic.com : from https://www.anthropic.com/news/claude-in-amazon-bedrock-fedr...
There's also their generic consumer ones, though I'd rate them as unlikely to do anything useful:
* support@anthropic.com : from https://www.anthropic.com/legal/consumer-terms
* privacy@anthropic.com : from https://www.anthropic.com/legal/privacy
And this out of left field one. They seem like actual lawyers:
* anthropicprivacy@bkl.co.kr : from https://www.anthropic.com/legal/privacy
---
Interestingly, Anthropic's "Trust Center" has an "Evidence of Insurance" document listed under "Other documents": https://trust.anthropic.com/resources#69eff53d22c228b34e5379...
Looks like you need to "Request Access", but if it's an automated system then it may give you access. And there _might_ be insurance contacts listed there who would be interested in this. :)
---
Follow up note -> Yep, it's automated and DOES give access to their docs. ;)
Their insurance levels don't really seem to be anywhere near what I'd expect frankly. To me, they look much lower than even entry level mandatory company insurance for brand new businesses at least in Australia. o_O
~~I mean, the worse part is the gif at the end of the message.~~
~~What are they even trying to do? What are they trying to convey? It just feels like being given the finger and getting my face rubbed in it on top of that.~~
Granted, it was very much weasel words.
Nonetheless, I read it as they were issuing a refund ("Let me look up your account information to help process your refund request."), but couldn't offer compensation for pain, suffering, loss of use, tracking down the bug, etc.
I could be wrong, of course, precisely because it was (probably AI-generated) weasel words.
We've been on this since the bug surfaced. Everyone affected is getting a full refund and an extra grant of usage credits equal to their monthly subscription as our apology. You can see my original post here: https://x.com/trq212/status/2048495545375990245. We’re still working on sending emails to everyone affected.
Our support flow wasn't set up to route a complex bug like this to engineering. We’re hoping to make this better but will take some time. Sorry to everyone caught up in it.
I got a random invoice for $45.08 back in March, despite not having auto top up enabled. Trying to reach support met with a brick wall. Based on the post I linked to, I'm not the only one facing this problem.
It happened this year to my one and only personal account. The account was one week old. Unique e-mail address. $5 balance for API credits. No usage yet. Suspended and refunded. Appeal denied without explanation.
I did create the account on a VPN because I was using public WiFi at a tech conference. That's probably what tripped their automation.
> I need to let you know that we are unable to issue compensation for degraded service or technical errors that result in incorrect billing routing.
What prevents you from issuing compensations?
Obviously someone can do it because it got done.
If the 'we' is referring to some team handling issues it would make more sense. In that case they should have said something along the lines of "I have informed someone who can help"
Would be more accurate. It still isn't setup. Talking to a bot as support who only tells you to talk to the bot for support is not actually support at all. It looks like support, but there's no way to ACTUALLY GET support.
For those of us not on X, what are the best communication channels for us to follow this sort of communication?
These fucks only respond when they get bad publicity.
> ugh sorry this was a bug with the 3rd party harness detection and how we pull git status into the system prompt
Claude wants to exercise control of how I use the "inclusive volume" that I purchased with my monthly subscription. This harms competition (someone else could write a more efficient or safer coding agent) and is generally not in the best interest of society. Why do we allow this?
This specific case is interesting, because it is so clear cut. There is no cross financing via ads, they already have the infrastructure to measure usage and even the infrastructure to bill extra usage. I also don't see how you can plausible make the argument that restricting usage to their blessed client is necessary for fair use or for the basic structure of their business model (this would be the standard argument for e.g. Youtube: Purposefully degrading the experience of their free client to not support background playback enables the subscription model).
I can’t use Claude Code online at all
Somebody (or something) wrote this code. This bug wouldn't be happening for any other reason. It's not a glitch, an oversight, a feature gap, or a temporary outage. It is a piece of written code in your system.
Everyone here is upset about the $200, which is probably much less money than the time that engineer spent ranting about the overcharge on GitHub.
The real problem in my mind is that that bit of code existed in the first place.
Why?
Are you vibe coding your billing!?
Without review!?!?
Or worse, a human being decided to add this to your code base? And nobody noticed or flagged it during code review?
Or much, much worse, Anthropic is purposefully ripping off customers?
This deserves a thorough post-mortem.
Happy to talk privately, but as I detailed here, https://news.ycombinator.com/item?id=47954005 . I've been billed $200 for a Max gift card to a 27 character alphanumeric icloud address that bounces.
I was looking through the system, and there are several UI/UX and process gaps in the gift card and billing order flow that expose Anthropic to significant liability. I'm genuinely not trying to concern troll or make some kind of overwrought threat here. Genuinely trying to be constructive. Let me give you an example.
I sent an email to Anthropic Support outlining the disputed / possibly malicious charge. The AI Agent / Claude instance agreed and replied with,
Thank you for confirming.
I've documented all the details about this unauthorized [specific amount + tax] charge for the Gift Max 20X subscription (invoice [lalala]) sent to [insert the random alphanumeric]@icloud.com.
An error occurred while evaluating the refund eligibility for your account. Your request has been fully documented and our team will follow up with you shortly to investigate this unauthorized transaction and assist with the refund and cancellation.
Best regards,
I'm not sure how familiar you are with international trading practises, but in multiple jurisdictions, the AI agent assumed legal liability for Anthropic. It accepted that the charge was unauthorized / fraudulent, stated that redressal was needed, but then failed to offer the means to redress it / didn't allow for the refund to continue.
I am not a lawyer, but based on my understanding of prior cases (I read this kind of stuff for fun, don't ask) – in the EU, the US and Canada, users can approach courts and invoke the doctrine of promissory estoppel (again don't quote me on this, IANAL, just like reading case law). And if enough users are affected / do so, it becomes a deceptive practises issue.
I've been thinking about how to solve this problem, and as strange as it sounds, I think Anthropic already has the tools to make the best customer support service in human history. No exaggeration. I think that this crisis could be an opportunity.
What does that even mean? Does it mean, "our support flow is just an LLM that fobs off customers and puts their issues into the bin"? Or is there some genuine "routing" of simple bugs to engineering which accidentally drops "complex" bugs? Could you drescibe that process, it sounds fascinating?
Also, how is changing a customer's billing based on detecting a certain string in a certain place a "complex" bug? Grep the string, remove the if statement, done. I'd love a post-mortem about why this was a complex bug.
More questions than answers here Thariq.
Please, please, please hire more humans with the actual ability to do the right thing for support if your AI agents can’t do the job.
Not sure I've ever seen a company openly take this position. This is a crazy policy.
And to get to that point, you need to be willing to spend a lot more than 200$.
A classic.
Back in December the iOS app had a bug ( https://status.claude.com/incidents/6rrnsb1y0kbn) in which buying a subscription thru the Apple App Store would not register with the backend, so you’d be charged but not receive the plan entitlement.
I discovered this because I wanted to upgrade from free plan to the regular plan. I was charged, but remained in the free tier. Thinking it was a temporary bug, I tried buying the max plan. Same result.
I tried cancelling the plan and restarting but I when I went to buy the regular plan again, I was forever tagged as an “Apple” user and so could only manage the billing plan on the iOS app. I tried one more time, same result.
I tried interacting with the support bot and although it agreed that there was a bug and that it should be fixed and I should get a refund, my account never was able to get unstuck nor refunded. I lodged a refund request with Apple, which was relatively quickly refunded. The Bot never did escalate to a human as promised.
Even though the bug was ostensibly fixed, my account (personal email) remains in permanent limbo, unable to upgrade from Free to anything else (I tried again recently and same result - paid but stuck on free plan). I had to create a new gmail just to pay for Anthropic / Claude.
Also when they added extra credits to everyone as an apology I was able to click the claim button multiple times and I got up to $400 in credits. Eventually a day later this dropped to $200 and then a few days later, $100 where it sits today.
I should have denied the PayPal charge on my bank account, that always gets a real human to look into it. Lesson learned.
He is getting a refund along with an additional $200 credit from what I can see.
It went massive on Reddit which is where I heard about it first.
Why would that vendor want to do business with a customer that doesn’t pay their bills (whether justified or not)?
Mismanaged comms? Yes
HN front page effect? Prob not
(could be Reddit frontpage effect or related tho)
I am cancelling my subscription as it is impossible to justify these degradations and paying for a subpar service especially now that we have at least 3 more models that are as good as Opus and there is the pi project that is undoubtedly the best harness.
In the last MONTH, I've asked how you can defend implementing (or even choose implementing) AI when:
the AI you have implemented throughout your company changes the results you've come to trust? https://www.theregister.com/2026/04/13/claude_outage_quality...
or won't let you log in?: https://github.com/anthropics/claude-code/issues/44257
or makes stuff up?: https://dwyer.co.za/static/claude-mixes-up-who-said-what-and...
or when it's down?: https://status.claude.com/incidents/6jd2m42f8mld
or when you get banned?: https://bannedbyanthropic.com/
or installs spyware: https://www.thatprivacyguy.com/blog/anthropic-spyware/
or takes the features you use out of the plan you subscribe to without notice? https://www.theregister.com/2026/04/22/anthropic_removes_cla...
or renders your IP legally unenforceable? https://legallayer.substack.com/p/who-owns-the-claude-code-w...
or stealthily changes pricing terms based on... file names you have? https://github.com/anthropics/claude-code/issues/53262
or invoices you for usage you did not perform, and won't answer support requests until you raise hell on social media? https://nickvecchioni.github.io/thoughts/2026/04/08/anthropi...
i mean seriously, why on earth would you use this? i thought we were professionals
On the other hand they make good products.
I'm also not sure if the person/bot who responded was saying "No refund" or that they couldn't issue a refund, or if a Github Issue was an appropriate place to ask for a refund.
Let's hope a human on the other end is reading this and acting accordingly. It all seems like we're only seeing part of a story.
Given that, it's almost guaranteed that sasha-id is a legitimate actor.
If you're confused about sasha-id's comment here (https://github.com/anthropics/claude-code/issues/53262#issue...), it's because they just copied and pasted a support response from Anthropic.
In this particular case I think the authors reply is them quoting what support told them?
It was obvious to me, but I can see how somebody could get confused from that.
I don't like it, but can't do much about it.
Is the culture really such that you can't escalate an obvious, fairly minor mistake that is turning into disastrous PR?
That would explain a lot of recent Anthropic takes actually.
- throwaway449933"Whether you think you can, or you think you can't—you're right" - Henry Ford
Sure, everyone probably has their own personal line such as "will quit if my employer is declared complicit in genocide by the UN", but bad customer service seems firmly in the "better to stay and advocate doing better from the inside" category
I don't see anything wrong with this. My integrity and values are above any company's. Companies can go to hell for all I care
if they can't do anything about it now, what makes you think that situation will change in the future? if remedial action would be punished by those higher on the ladder, it certainly won't be promoted by those folks, leaving this hypothetical employee in exactly the same position they're currently in.
quit.
So basically all of big tech.
How about Anthropic agreeing to a $1.5 billion settlement for perhaps the biggest theft in history?
Weird how people forgot about that.
An IC won't be able to steer a ship like that back to morality. Whole teams can't do it. People at Google organized to stop this sort of shit and were fired IIRC?
Large institutions provide cover for bad actions by people who, without said cover, would not take those actions.
Therefore, I believe that "we'd be left with only people who are cynical and/or bad and/or sufficiently indentured to be unable to push back against management, and there would be no hope of the company ever improving" is the status quo.
So what are you even saying??
I had organizations leaders before say things that are so black and white like "We should delete all user accounts that haven't logged in 6 months", you say "Are you sure? some people will be upset. Some will post on twitter or reddit and complain etc" they confidently reply "Yes, we will explain that it's not sustainable and they are welcome to create another account". So you go ahead and implement that. 1 second after it goes into effect, you get angry support tickets, a post on twitter, and that "leader" immediately backpedals that "the implementation was not how I expected". Like what did you expect was gonna happen exactly?
But my best guess is they don't want to put a firm line down because they want to be free to shift it around however they'd like.
At the same time, it's clear that after this happened, Anthropic took action. 3 DAYS AGO! (https://news.ycombinator.com/item?id=47954655)
That's before this comment was made on the issue:
https://github.com/anthropics/claude-code/issues/53262#issue...
I'm surprised Anthropic didn't also say this on the issue. Weird that they wouldn't. It seems to have made for unnecessary bad PR.
It feels to me that Anthropic is less focused on quality, and more focused on PR stunts/flash. My experience with Claude is always "it's pretty and feels cool", where-as codex feels like "solid and boring". I realize I'm probably biased. Am I alone in this thinking?
He's getting a refund + $200 worth of credits
Meanwhile I've integrated CC into my workflow enough that I'd feel frustrated cutting out all LLM agent use.
I don't have the hardware to run models locally, and I'm not excited about the idea of spending that money. I could use a different harness with one of the services that runs open-weight models for me, but I feel like the cost would be prohibitive. I'm paying $100/mo right now and that's all I'm willing to spend.
Personally tried GLM subscription. Bought it during new years discount: 36$ for a YEAR.
Cannot burn tokens through with personal project use. From what I can see in stats they allow 25-100M tokens in 5h period (for cheapest plan), depending on the model. GLM5.1 could be a bit slower and likes to (over)think, but I don't see practical differences from Sonnet 4.6 or Opus 4.6.
> I refuse to give Altman money, but Anthropic keeps disappointing me over and over with crap like this. Gemini seems behind? Not touching Grok.
My thought process is totally the same. And even there's slight concern about ethics using GLM, at least in my conciousness, openai is worse and grok is the worst of them all by far, no competition.
This is the most interesting line to me. "Anti-abuse system"? I would bet the system is far from being just a conditional on a specific filename. In other words, this supposed anti-abuse system might be far more pervasive without the user's knowledge. And perhaps even more importantly, who thought upcharging instead of blocking is the correct approach to dealing with this alleged "abuse"? Is this some anti-distillation feature they let Claude itself write looking at past distillation attempts producing similar artifacts or what?
Credit card didn’t get through, pro plan got insta cancelled, had to pay for full max plan. Clearly a billing bug on their side. If the credit card when upgrading a plan doesn’t come through, don’t destroy the existing plan.
I talked to the chat bot; i got a ticket number, a human will come back to me. That was three months ago. Never got refunded. Nobody emailed me.
I ended cancelling the max plan, it expired yesterday. This plus the constant degradation of the service despite having 30B revenue first quarter this year.
A company that has so much money, and cannot care less about their users…
They will have to do much better if they want to get me back.
When you've been a Software Engineer for a while you start to be able to put bugs in certain buckets.
Then there is the last bucket, like the X-Files. They don't belong anywhere else. They have no specific reason. They happened because of a weird set of circumstances, usually due to too many developers working on the same product, without proper abstractions and separations.
And having spent too much time that I'd like working and reviewing code generated by AI, this is exactly what the AI does. It doesn't abstract. It doesn't separate. It just does what it is asked, not that different from the quality of code from outsourcing contractors.
I would have jumped ship, but OpenAI saying "hold my beer" when Anthropic declined the Pentagon's safeguard removal demands is the only thing that has prevented me from jumping ship. I've considered Chinese AI services but I'm too concerned with data (proprietary code) exfiltration.
For example, chatgpt when asked "How to report a billing issue with Anthropic subscription?" says:
Best way: Use Claude’s built-in support Log in to your Claude account at Anthropic / Claude.ai Click your initials or name in the lower-left corner Select “Get help” Use the support messenger to describe your billing issue (duplicate charge, failed renewal, refund request, missing credits, invoice issue, etc.)
I got a $2 charge for a Facebook Ad (I know, $2 is nothing and I shouldn't use Meta), and it was completely wrong. It's impossible to talk to someone in Facebook about this. The AI chat is completely clueless and can't do anything. Their help page say you can ask for a refund (I can't, because the payment doesn't appear on the billing page or payment activity), but they tell you they will close your account if you do it, like... wtf?
I am scared for the future where AI handles all of this. It should be ilegal. Companies should have a X support people every Y customers or something like that. I see it everyday and it's getting worse and worse...
Some days I think the only solution is what Bombita did in the movie Relatos Salvajes https://www.youtube.com/watch?v=vP3IwmM3XLQ
There should have been a two layer approach:
1. Run regex to screen for target word
2. If positive, run the context through a cheap AI model
you will do it again because you are an all-day sucker
Just refuse to pay any bill from any vendor that by their own public admission) is a "incorrect bill".
This isn't just about PR and technicalities, this is Business 101.
English is not my first language, so I might have misunderstood....
The rest of the support response is just pleasantries and padding, to dance around this fact ("Your detailed reproduction steps will be valuable" blah blah).
What a claude excuse
(Virtual card provider that generates cards as a free-to-the-user service. They make their money from a cut of the standard transaction fees. Cards are locked to a single merchant and it’s easy to configure limits.)
Anthropic will need to make sure that i am never charged beyond my subscription fees before I consider a sub.
A. Do shady shit to try to hurt competitors/stop distillation/stop pro plan users using other coding harnesses/etc.
B. Any shady shit that gets viral enough gets a PR apology response and removed.
C. Add new shady shit. GoTo => A.
And I am incredulous as to how this opinion is not more widespread.
The person who created the PR is user "sasha-id".
The person saying no to the refund is also user "sasha-id".
What?
Where was it exactly thats someone from Anthropic said no to a refund request? I feel I am missing the obvious somehow.
He didn't quote the Anthropic response, leading to the impression that he was Anthropic staff, confusing you, me and "CollectionAgency" in the issue thread, among others, I assume.
No response from customer service.. only their AI Agent Support.. Which has still not offered me a refund.
I may have to do a chargeback.
I don't think it's as sinister as you're implying. I think it's part of them disallowing 3rd party clients from using Claude Code subscription and someone making a bad assumption that certain files in a repo being a good signal that someone is attempting to bypass those rules.
It's still not a good look for Anthropic, but I don't take this as a secret attempt to sabotage a competitor. I take it as them trying to enforce rules that they had very publicly announced.
Ah yes, cause who bothers to test any releases to actual paying customers
I think people put this out of proportion. Yes, you can reason this is ethically correct - I don't object to this. But people used Anthropic, Claude etc... in the first place. Why would you use something to then be disappointed about how it performs, when it comes to AI? Would not be the better and easier strategy to ... not use it in the first place, and make yourself dependable on AI? I don't fully understand this. I would not run into a similar situation because I simply don't use any AI. I actively want to support those folks who don't use AI either - that way we can point out all the ill effects of AI, such as in the case of Anthropic to prioritize on greed.
Allow users to file a lawsuit against the company using AI against their customers and judge the company only on what the AI generated without a chance to add anything more in their defense. Also any boilerplate legalese the AIs will quote in reaction to such laws is null and void.
Suddenly every AI support channel will have an "escalate to human support" button.
I can’t believe they paid 100m for some of these employees. They could have bought entire companies of real developers.
The deeper into the new world order, the more you'll be charged for every breath, by design and by bugs-as-features all the same, refunds be against technofascist manifestos.
On other hand I wonder what other filenames one could include in their repos to cause this sort of behaviour. Kinda a nudge towards people leaving these tools.
Krasis is one such tool that allows large models using blended GPU/RAM.
ik_llama for better performance than llama.
ComfyAI for local image generation.
Nanocrab seems better for orchestration. Still need a good system capability firewall.
https://github.com/anthropics/claude-code/issues/53262#issue...