New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Bruh is sus af, Claude clock it as opp from its j-space
One of the automated ban triggers is using 100% of your 5h/1w quota 100% of the time
不要浪费时间了，这里没人会帮你，实际上，在中国使用Claude也不是一件非常困难的事，方法有不少，只是你搞不懂而已。
我给你指条路吧，你到大城市去租一间高档写字楼的办公室，有一些写字楼的网络是登记在外资企业名下的，使用他们的网络和IP连接Claude没有任何限制，包括订阅也不会出任何问题
--------gemini translate--------
"Stop wasting your time. Nobody's going to help you here. In fact, using Claude in China isn't that difficult—there are plenty of ways to do it, you just haven't figured it out yet.
Let me point you in the right direction: go rent an office in a high-end commercial building in a major city. Some of these buildings have their network registered under foreign enterprise accounts. Using their network and IP gives you unrestricted access to Claude, and you won't run into any issues with subscriptions either."
If you're from the U.S., have you tried opening a port and using Python to create a tunnel so that your VPS's IP address uses your residential IP address?...
Alternatively, you could buy proxies that offer residential IP addresses and create a tunnel to those residential IP addresses.
Unfortunately, AI learns all the shortcuts that are posted online
it's far more aggressive than that right now in China if we are talking about Anthropic, OpenAI is fine
sometimes I think Anthropic's ban on users in China is actually just because they lack computing capacity, national security is purely an excuse
yeseses, that just one of them. Average US user will make their thing and leave % of quota unused. Average CN user uses 100% of quota, if they have left over before reset, burn it on useless things or sell remaining capacity.
Other triggers include prompt content analysis and timezone of use matching billing address, I have seen bans on my transfer station for asking to write discuz plugin, because not single US legitimate billing address will do that
haha, don't use China only script
and wait, so you can see their prompt and history, right?
Yeah, I do not actually run airport or train station or whatever, just have a large pool I allow some friends to use, since have maximum plans for all providers . (all non-CN though)
I think one of biggest reason that people want to go in Claude/OAI directly is that the relay cannot be trusted, since they terminate similar to cloudflare and re-send the request, many people do not understand that the relay station can also send anything they want, that includes telling your client to run a tool call to delete everything on disk, send contents of secrets. I dont understand why people think the biggest risk is being resold cheaper model, the relay can shell command or RCE anything on your device if you run in auto mode as a response to a chat completion
The other ban that will re-apply to you repeatedly if you keep losing accounts.. you cant REUSE the same repos or projects. When you get banned for TOS, for being chinese, or similar, many people will resume working on same repo / resume session after swapping out backend or keys or account... that is flagged automatically as same session unique IDs re-joining the banned user and the new account
Simply ask the AI to hack itself to grant you access without IP restrictions.
I think you should chat with AI more often.
I constantly use the OVH / Hetzner data centers, and I don’t have any problems. Claude/OpenAI doesn’t ban me. The problem is that you’re not banned for your IP as DC. They simply determine it based on the language of prompts
lmao
The biggest issue concerns the registration and use of accounts, as well as credit cards—these are flagged not when they are being used, but right from the moment registration begins.
Why do you use AI so much? That's a lot of money!
I don't use AI at all, so I am a bit behind on the times. A lot of people seem to claim that the Chinese models are as good as OpenAPI+Anthropic, but then it seems like many Chinese want to use the American models. Why is that?
It seems like if no-one used AI, a lot of problems would go away. Progress would be "slower," but is that so bad?
Or learning to code.
Buddy, if you were creating something of value, those subscriptions would have multiple times payback in revenue. Wtf are you actually doing besides illegal stuff?
If you can waste $1100/m on ai subscriptions and complain you're doing something wrong.
Either they are doing something wildly wildly illegal or whoever they are paying for credit cards is stealing the refund from them too. Anthropic refunds you when they terminate your account for TOS violations, this was when they canceled my account for asking it to disassemble a binary:
If you are using a legit card that is yours, even if its an overseas card, you get your refund. If you are paying a criminal for a card number to use then . . .
It's cost-benefit analysis type of equilibrium gated, such in when you are doing "major" work, the Chinese pro models are behind, but the Chinese flash (fast, low cost, resource) model is ahead.
The cost best ratio is for some (undefined as of yet, they are all equalvalent) American frontier modern expensive model to act as orchestration layer, to spawn and instruct subagents running low cost Chinese Flash models to do actual work in very small slices.
Having Chinese model do the orchestration and review is generally either waste of tokens or disaster of code. Having American frontier model (ex. Fable or Astra) write entire code is expensive as hell. If you have a simple project yes deepseek pro is fine, it will one shot a python script fine for 5 cents compared to $10 on claude.
I don't like the way people are going full AI psychosis like it will take over humanity, but I work at company that now makes frontier AI and have to know this shit against my will
I hope this was a joke.
How do you know OP is not a North Korean plant?
We offer mini PCs with residential internet connections in Vietnam for rent, in case anyone needs them.
They canceled your fucking account for asking it to disassemble a binary? Holy mother of censorship.
I think I have one or two servers that are marked by IPInfo as being a fixed-line ISP. If anyone wants to use it to forward their connections to some API and has a legitimate need (and isn't brand new to LET), DM me and I'll see what I can do.
Yeah I've been doing it with specific tools for a while without a problem, but trying the recent rea skill (reverse-engineer-anything) that went "viral" immediately had account terminated for abuse and attacks.
Should have try to decoompile first
It seems that risk control in Russia isn't all that strict.
You want billing address and IP that is close to your time zone, it does not matter, if your peak usage hours are in UTC+8 base line and you are using the cleanest residential USA IP ever it does not matter, you are detected.
That’s why I say no provider is willing to sell this type of VPS; I’m accessing it remotely via a paid subscription, but their perspective on the matter is a bit off-base.
The SNI-restricted residential egress idea is technically sound — it's basically what many corporate proxies already do for SaaS whitelisting. The real blocker, as karcode0 hinted, is IP supply: clean residential IPs are finite, and the moment one lands on a denylist the product's value collapses, so a host would need constant churn and re-verification. Honestly the economics probably only work if it's priced well above a typical VPS, which clashes with what this crowd expects to pay.
I don’t get it