Approach to continuing using A/ models
What matters is the model; the gap between harnesses isn’t that big. Recently A/ has opened up some reverse-proxy access, so it’s worth trying other harness tools. opencode, copilot cli, etc. are all good.
What matters is the model; the gap between harnesses isn’t that big. Recently A/ has opened up some reverse-proxy access, so it’s worth trying other harness tools. opencode, copilot cli, etc. are all good.
I used to most recommend chatgpt plus, but the other day chatgpt banned my codex. Logging into codex again requires SMS two-factor authentication. Since it was a number from a code-receiving platform, openai does not support self-service phone number rebinding, so the main account is basically dead. I switched to try out supergrok, and the composer it provides…
The second screenshot is enough to prove it’s genuine, the familiar 1302 flavor. Stable to use on 6.18, speed is quite good, maybe the supplier no longer uses z.ai. In practice, it’s stronger than deepseekv4pro in coding, script error rate dropped a lot. Of course, the price is also many times higher. Among domestic models, only deepseek is recommended, hoping it soon…
Requires configuring scheduled tasks on a computer; you can look into it yourself
Not cheap, especially the way token consumption is counted—both input and output are counted within the plan. I often use 20 million tokens in a single request, and the 59 yuan tier only covers about 3 of my requests. For coding, roughly 99% of input is input cache, and billing cache hits at full price is pretty absurd. If they want people to actually use it, they shouldn’t count input—only count output toward the quota,…
2.5折 directly becomes 1/4 permanently, no need to rush purchase, no packages, no need to grind 5h daily or weekly quota, no need to calculate off-peak/peak multipliers, no need to calculate token-to-credit exchange rates, direct retail equals wholesale, no tricks, all sincerity.
My main use case is writing code, and my projects are mainly command-line tools. Tencent HY3 Preview performs very well in my use case, with good instruction following and dialectical logic. I also tried Baidu’s CoBuddy coding model recently launched on openrouter, and it’s terrible—just two simple uses reveal its poor capability, with only 131k context…
Archive of a Xiaohongshu short post about using a simple prompt in ChatGPT to check your own usage statistics.
Got a refund again, freeing up resources for the bros. 40 bucks a month is indeed cheap, but it’s unusable during the day and I need to sleep at night—this isn’t a lite plan, it’s a night plan! Refund survey, zhipu-ai.feishu.cn/share/base/form/shrcn8hoW11cnI0PRd7H…
This is the third time I’ve been fooled by Zhipu. They said they were retaining old users by giving unlimited weekly quota, so I came back to give it another try, and got tricked again. It has no weekly limit, but it’s completely unusable when I need it. It only comes back to life right when I start and finish work, which is like meeting at the peak. In February I said Zhipu lite had single concurrency, and Zhipu used its business certificate to report me to xhs…
Archive of a Xiaohongshu short post about ChatGPT Free Account Benefits End.
On Claude’s side, max 5x costs five times the pro price with five times the usage, and max 20x costs ten times the price with 20 times the usage. On OpenAI’s side, pro is six times plus, and review is ten times, making the cost-performance hard to judge. Of course it also has the privilege of prioritized request processing, and I’m not sure how much faster it can be. I use chat for daily conversations, with extend thinking on, and it feels like an average of about five minutes per question. Plus has a weekly limit of 3000 conversations, which is far from being used up. The codex cli weekly quota is heavily enough for two days (without fast mode on), so you only need to prepare two or three plus accounts to meet weekly needs. If you’re impatient and turn on fast mode, having four accounts ready is also enough.
At peak I subscribed to several plans at the same time, but couldn’t use them all up, wasting at least half of the few hundred bucks every month. OpenAI recently opened codex cli to free users with a certain free weekly quota. Those familiar with codex know its quota is very different from claudecode’s, and is extremely durable. I myself have 6 accounts connected behind a cli proxy, and using only the free tier already basically meets my needs. Some share that free accounts may have model quantization and reduced intelligence; I also have an account in an enterprise team, and later I’ll carefully test whether free GPT-5.3-codex differs from the paid one.
The refund window will open within one week after the holiday, and close on March 6, 2026. At that time, go to the My Plans page and click Cancel and Refund to complete it with one click. Note that the refund window period ends on 3.6. Programmer friends, recall your experience using Zhipu in the past two months, trust your own feelings, and make a rational decision.
Many people say the lite plan has low concurrency, but they may not have made clear exactly how low it is. Let me explain it more clearly. Only one concurrency. This means in tools with a subagent mechanism, lite simply cannot run. I have been buying various AI products for four or five years, and Zhipu’s plan is the biggest rip-off I have encountered; it is almost unusable for coding. Now it can only be used for single-thread…
After recharging $10 on OpenRouter, you get 1000 free model calls per day, which is more than enough for me to use for validation. But for a long time, many of its models kept returning 429 errors, such as openai/gpt-oss-120b:free. Today I manually tested various methods and finally confirmed…
Open-source repo github:jqknono/reference-Harmony Main features see image, displays shortcuts and common commands for various popular dev tools, which help improve work efficiency. “Dev Cheatsheet” has clear architecture, simple and focused features, will never show ads in the future, and fully complies with Huawei’s app review policy. Submitted in Dec…
The Qwen app uses four-direction swipe gestures. On the left are chat history and agents, and on the right is the agent marketplace. When the swipe trajectory on the screen is not straight up and down, but rather like a left-falling stroke (丿) or right-falling stroke, the side content will be triggered. To return to the chat interface, you don’t swipe the screen, but perform a back gesture, sliding in from the screen edge to return. If the directions get mixed up, the app will freeze. The Qwen app…
Inconvenient to start a new conversation; after starting one, only a popup notification appears while the old messages remain in place. Even after switching topics, the previous answer is still visible, giving a feeling that the context hasn’t switched—which is uncommon among AI chat apps. However, Doubao’s lower barrier to entry does help it grow its user base.
Archive of a Xiaohongshu short post about Zhipu strict concurrency limit.
Archive of a Xiaohongshu short post about sharing a cost-effective service at Tongji for pregnancy preparation.
Archive of a Xiaohongshu short post: DeepSeek v3.2 works better with claude code.
Archive of a Xiaohongshu short post about my first computer with 16MB RAM.
Archive of a Xiaohongshu short post about ModelScope now limiting single model to only 100 daily uses.
Archive of a Xiaohongshu short post about Alibaba Cloud ESA being able to forward DoH.
Archive of a Xiaohongshu short post sharing a mouse button remapping setup suitable for programmers.
Archive of a Xiaohongshu short post about WeChat’s minor mode being able to block ads.
Archive of a Xiaohongshu short post about Don’t use cursoragent review.
Archive of a short Xiaohongshu post: Please Learn to Cook, Okay?.
Archive of a Xiaohongshu short post about HarmonyOS version of WeChat Reading offering free book reading.
Archive of a Xiaohongshu short post about no sweet-spot local LLM hardware having emerged yet.
Archive of a Xiaohongshu short post: Prioritize Using the Most Advanced Models to Save Time.
Archive of a Xiaohongshu short post about Zhipu’s plan being worth a try.
Archive of a Xiaohongshu short post about gpt-5-codex实测不如gpt-5.
Archive of a Xiaohongshu short post about Kimi K2 cost-performance being mediocre.
Archive of a Xiaohongshu short post about powerful personal AI hardware being both a helper and a shackle.
Archive of a short Xiaohongshu post about As expected, HarmonyOS system-level ads are back.
Archive of a Xiaohongshu short post about sharing on the working style of large models.
Archive of a Xiaohongshu short post about claude code alternative.
Archive of a Xiaohongshu short post recommending a completely free coding tool roo code.
Archive of a Xiaohongshu short post about Github Copilot is outdated.
Archive of a Xiaohongshu short post about Myths about Gemini Code Assistant Membership.
Archive of a Xiaohongshu short post about hoping domestic AIs can all offer a package plan.
Archive of a Xiaohongshu short post about DeepSeek supporting Claude Code.
Archive of a Xiaohongshu short post: A blunder by Cursor after the launch of GPT-5.
Archive of a Xiaohongshu short post about only God can understand vibe coding.
Archive of a Xiaohongshu short post recommending a practical AI coding package combo.
Archive of a Xiaohongshu short post recommending AI coding tools with great experience so far.
Archive of a Xiaohongshu short post: A few fair words for qwen3 code.
Archive of a Xiaohongshu short post about Do Not Ignore the Potential Risks of Claude Code Relay.
Archive of a Xiaohongshu short post about Mate60Pro no longer overheating after upgrading to 5.0.
Archive of a Xiaohongshu short post: Qwen App’s understanding of the product is like tarot cards.
Archive of a Xiaohongshu short post recommending Huawei phone users to update the new version of Smart Search.
Archive of a Xiaohongshu short post on why to close public DoH.
Archive of a short Xiaohongshu post about Manus should not rush to launch.
Archive of a short Xiaohongshu post: Has anyone used the free Grok API?
Archive of a Xiaohongshu short post about whether you should give your child a separate mobile phone.
Archive of a Xiaohongshu short post about WeChat Reading auto check-in.
Archive of a Xiaohongshu short post: Set up an ad-blocking service in 15 seconds.
Archive of a Xiaohongshu short post about a computer that can only learn programming, not play games.
Archive of a Xiaohongshu short post about AdGuardPrivate–Publicly Accessible AdguardHome.
Archive of a Xiaohongshu short post about Ad Innovation: Wipe-to-Jump.
Archive of a Xiaohongshu short post about Ad Innovation: Shake-to-Jump.
Archive of a Xiaohongshu short post on Beware of Contract-based Billing for Video Memberships.
Archive of a Xiaohongshu short post about Monthly Challenge Reading Membership Is More Cost-Effective Than Expected.
Archive of a Xiaohongshu short post about a classic bad computer habit that got hacked.
Archive of a Xiaohongshu short post about fewer books available for WeChat Reading trial membership.
Archive of a Xiaohongshu short post about Meituan on Huawei phone continuously collecting information.
Archive of a Xiaohongshu short post about Optimal Strategy Analysis of WeChat Reading.
Archive of a short Xiaohongshu post about Don’t Believe Huawei Ads Can Be Turned Off.
Archive of a Xiaohongshu short post about how to install Google framework on Huawei.
Archive of a Xiaohongshu short post about adding a free public service supporting DNS pollution prevention.
Archive of a Xiaohongshu short post about why Copilot cannot be used.
Archive of a Xiaohongshu short post: My self-hosted ad-blocking service, welcome to try.
Archive of a Xiaohongshu short post about Huawei phone supports replacing Google suite minimal method.
Archive of a Xiaohongshu short post about Mate 60 Series Major Update: Disable Call Ringtone.
Archive of a Xiaohongshu short post about what’s wrong with Alipay adding a video feature.
Archive of a Xiaohongshu short post about two methods to create WeChat Reading auto check-in tasks.
Archive of a Xiaohongshu short post about auto check-in to WeChat Reading using Huawei Smart Scene.
Archive of a Xiaohongshu short post about blocking all ads on Android phone with one setting.
Archive of a Xiaohongshu short post about truly disabling Huawei pull-down search ads.
Archive of a Xiaohongshu short post analyzing why Huawei Smart Search insists on slacking off.
Archive of a Xiaohongshu short post: Copilot excels at formal logic, just like some junior programmers.
Archive of a Xiaohongshu short post about Mate 60 Pro still at 100% maximum capacity after half a year.
Archive of a short Xiaohongshu post: Do I have autism?
Archive of a Xiaohongshu short post about comforting my wife made me anxious too.
Archive of a Xiaohongshu short post about Huawei System App Experience - Weather.
Archive of a Xiaohongshu short post about Huawei and Zhihu’s In-depth Strong Alliance.
Archive of a Xiaohongshu short post about Huawei and Meituan’s In-depth Strong Alliance.
Archive of a Xiaohongshu short post: Huawei thought it over at 2:30 AM and finally gave up on ordering me late-night snacks.
Archive of a Xiaohongshu short post about Huawei always wanting to lend me some money.
Archive of a short Xiaohongshu post about why HMSCore always launches Meituan.
Archive of a Xiaohongshu short post about Huawei Browser’s Search Engine Ads.
Archive of a Xiaohongshu short post about how the main problem with Huawei is the lack of dignity that only users can perceive.
Archive of a Xiaohongshu short post about Mate60Pro Poor Photo Solution–Pro Mode.
Archive of a Xiaohongshu short post about Mate60Pro Overheating and Short Battery Life Solution–Power Saving Mode.
Archive of a Xiaohongshu short post about basic experience sharing after switching to Huawei Mate60Pro.
Archive of the Xiaohongshu short post ‘My wife asked me why I didn’t go play ball with them, and she was stunned within five seconds’.
Archive of short posts from Xiaohongshu account 紫气喵来.
Archive of the Xiaohongshu short post ‘Precautions for Climbing Steep Mountains’.
Archive of the Xiaohongshu short post ‘Heaven knows, earth knows, you know, I know, gods and ghosts know—what is there left to be unknown?’
Archive of the Xiaohongshu short post ‘Follow scattered families to avoid looking lonely’.
Archive of the Xiaohongshu short post ‘The Anlu Expressway Service Area Is Beautifully Built’.