Yeliang

浮华公益站

[Yeliang @Yeliangcloud]

@wyl2035 DONE

[Yeliang @Yeliangcloud]

@davezfr @RonVonng 所以你的8＄购权是无效的，得用20＄购权，不然你这样会让很多人误以为只需要购买8＄的go套餐就可以了

[Yeliang @Yeliangcloud]

@thsottiaux Why was I banned today after topping up my Pro20X account yesterday? I spent over a thousand yuan, I'm very upset, and I'm planning to switch to Claude.This is my email address: [email protected]

[Yeliang @Yeliangcloud]

@thsottiaux @thsottiaux

[Yeliang @Yeliangcloud]

@thsottiaux Why was I banned today after topping up my Pro20X account yesterday? I spent over a thousand yuan, I'm very upset, and I'm planning to switch to Claude.This is my email address: [email protected]

[Yeliang @Yeliangcloud]

@thsottiaux Why was I banned today after topping up my Pro20X account yesterday? I spent over a thousand yuan, I'm very upset, and I'm planning to switch to Claude.This is my email address: [email protected]

[Yeliang @Yeliangcloud]

@mjjwiki 新加坡IP＋Starrybui卡可以

[Yeliang @Yeliangcloud]

@derrickcchoi 当前codex配置Cline的deepseek-v4-flash模型时，容易出现乱码、无限制输出；Windows最新版本codex的MCP超过10个会出现疯狂��取硬盘，写盘，日志剧增；对于Windows的Power shell，目前终端没有做“自动”适配选项，我认为可以增加“自动”适配，即让模型自主选择Git Bash（远端运维）或Power shell（本机）

[Yeliang @Yeliangcloud]

@thsottiaux Why is it that with my Plus plan—which started with 100% of the quota—sending a single "hi" caused a 27% drop? It’s very strange; this issue hasn't been fixed since the reset last week.

[Yeliang @Yeliangcloud]

@cline When will you be able to integrate the official version of deepseek-v4-flash? I've already purchased a ten-year subscription from you.

[Yeliang @Yeliangcloud]

推荐一个接码平台：SMS Pool
如果你���像我一样，喜欢用未接码GPT半成品号，那么这家将会是一个很不错的选择，因为它有如下优点：
①实卡号码，接码准确率高
②特别适合用来接GPT，价格便宜，部分国家哪怕是到WhatsApp验证码也能接收
③充值很方便，支持信用卡/借记卡/支付宝/加密货币（推荐用加密货币，不限制最低充值金额）
④号码保留六天，期间不限制任意收取验证码
平台地址：https://t.co/nJq42vwOHQ

Here is a recommended SMS verification service: SMS Pool.
If, like me, you prefer using "semi-finished" ChatGPT accounts (accounts created via SMS verification services), this provider is an excellent choice for the following reasons:
① Uses real SIM card numbers, ensuring high success rates for receiving codes.
② Ideally suited for ChatGPT verification and very affordable; for some countries, it can even receive verification codes for services like WhatsApp.
③ Convenient top-up options, supporting credit/debit cards, Alipay, and cryptocurrency (cryptocurrency is recommended as there is no minimum deposit requirement).
④ The number is reserved for six days, during which there are no limits on receiving verification codes.

Platform URL:https://t.co/nJq42vwOHQ

[Yeliang @Yeliangcloud]

@GaoSteven2010 GPT短链支付需要支付卡和支付国家（家宽）一致才可以，转出长链或者Google Pay或者Apple Pay这些就可以直接过，我都订阅过

[Yeliang @Yeliangcloud]

体验更自由的 AI 订阅方式——BitMart 预付借记卡，轻松完成 快速 KYC 即可开卡。
专属权益：一键订阅 ChatGPT、Claude 等主流 AI 服务；无限消费、随时充值；支持全球通行，随时随地使用。
立即申请：https://t.co/UTMTqlSJ2q

[Yeliang @Yeliangcloud]

@QinshitD Cliproxy 10G流量 35元

[Yeliang @Yeliangcloud]

Why I switched from Claude to GPT-5.6 Sol — and what I genuinely love about it
For the past few months I’ve been a heavy Claude user, especially the Fable series. It was excellent for long-form writing, deep reasoning, and careful code review. But recently I kept hitting the same frustrations: quotas disappearing way too fast, sudden refusals or overly cautious responses, and agentic workflows that would drift or give up halfway through complex tasks.
Then I decided to seriously try GPT-5.6 Sol, especially through Codex.
The first thing that hit me is how much it feels like working with a real engineer who actually finishes the job, rather than just a smart chatbot.
In coding tasks, Sol blew me away. I fed it a medium-sized project I had been iterating on with Claude for days (around 15k lines of Python involving data pipelines, API integrations, and frontend interactions). It quickly grasped the overall architecture, then went far beyond just writing code — it ran verification, wrote tests, caught edge cases, and kept refining until everything actually ran smoothly. That “get it done” tenacity is something I rarely saw before.
Token efficiency is another huge win. The same tasks often require noticeably fewer tokens while delivering equal or better results. This means I can do way more work within the same budget, or run much more ambitious agent workflows for less money. Paired with Codex, it turns “writing code” into true collaborative development.
The agentic capabilities stand out too. Ultra mode lets it coordinate multiple sub-agents in parallel for complex jobs. When I need to research, code, test, document, and deploy all at once, it stays on track and adapts strategies mid-process. For long-horizon tasks, it doesn’t lose the plot — something that used to break my flow constantly.
Another practical advantage: far fewer unnecessary refusals and guardrails. Claude would sometimes block or over-complicate boundary cases, while Sol is much more willing to solve the actual problem (while staying safe, of course). That “usability” boost has a massive impact on daily productivity.
I’ve also been impressed with its strengths in knowledge work, design judgment, and scientific tasks. It helped me analyze bioinformatics workflows and generate clean frontend prototypes — outputs were high-quality, well-structured, and immediately usable.
Of course it’s not perfect. In extremely long contexts I still need to guide it occasionally, and sub-agent configuration could be smoother. But overall, the experience was strong enough that I’ve made Sol + Codex my daily driver.
One more thing I really appreciate is the transparency and responsiveness from the OpenAI team. Tibo regularly replies to feedback on X, explains changes, and even resets limits when needed. That direct communication makes me feel heard and more willing to invest deeply in the platform.
Here’s why I switched and why I love GPT-5.6 Sol:
It actually completes complex coding and agent tasks instead of leaving half-finished work.
Outstanding token efficiency and better performance per dollar.
Fewer refusals and higher real-world usability.
Seamless integration with Codex for a polished workflow.
Much more stable on long-horizon, multi-step projects.
A team that listens and communicates openly.
If you’re on the fence or coming from another model, I strongly recommend jumping in and building something real with it. Theory is one thing — running a full project is where the difference becomes obvious.
Huge thanks to Tibo and the OpenAI team for shipping this model and running this campaign so more people can experience it.@thsottiaux

[Tibo @thsottiaux]

Or… what if we gave you $100 in Codex credits if you tell us what you love about GPT-5.6 Sol or why you switched?

Tweet it, claim your gift, enjoy more usage. First 10k get the free tokens!

https://t.co/8mU93eA13i

[Yeliang @Yeliangcloud]

@Cryptoxiaodd 贵了，已经有1元的货流出了，2个起售

[Yeliang @Yeliangcloud]

@ModengSir 现在gemini AI Pro 一年都8元直充了