博文

目前显示的是 十一月, 2026的博文

AI made cover letters better, and that's the problem. How to write one that stands out

图片
Here's an awkward finding. A 2026 study from Tilburg University, reported by IO+, found that AI does improve cover letters, but barely improves the chances of being invited to interview. When everyone polishes their letter with the same tools, the letters start to sound alike. Hiring managers notice. In a CV Genius survey of 625 hiring managers, 80% said they view obviously AI-generated CVs and cover letters negatively, and 74% said they can spot them. Yet cover letters still matter. Most hiring managers say they read them when they're submitted, and nearly half say a strong one can win an interview for a candidate with a weaker resume. So the goal isn't to avoid AI. It's to avoid the letter everyone else is sending. Three steps to a letter only you could write Give the AI the job posting and ask it to pick out the three requirements that matter most. Match each one with something you actually did , with a number if you have one. Only you can supply this part....

投外企、接海外单要写 Cover Letter?用 AI 写,但别写成"人人都一样"的那封

图片
国内找工作很少要求求职信,但投外企、去海外求职、在 Upwork 这类平台接单,Cover Letter 几乎躲不开。V2EX 上就有人因为在 Upwork "经常要写 cover letter",干脆自己做了个自动生成的网站。 用 AI 写当然快。问题是,看信的人怎么想? 招聘方的真实态度 CV Genius 调查了 625 位招聘经理:80% 对明显是 AI 生成的简历和求职信印象不好,74% 说自己看得出来。 荷兰蒂尔堡大学 2026 年的研究发现:AI 确实能把求职信写得更好,但几乎没有提高拿到面试的机会。大家都用 AI,信就越来越像。 但求职信并没有过时:多数招聘经理说,求职信交了他们大都会看;将近一半人说,一封好信能让简历稍弱的人拿到面试。 结论很清楚: 能用 AI,但要写出只有你能写的东西。 3 步写出不"撞脸"的求职信 把招聘启事原文给 AI ,让它挑出最重要的 3 个要求。 每个要求配一件你真实做过的事 ,有数字就写数字。这一步只能你自己给素材。 要求 AI 不许编、不许套话 ,写完你再读一遍:像不像你会说的话?开头那句客套话,通常可以直接删。 可直接复制的提示词(中英文都能写): 用免费 AI 写一封求职信 参考: V2EX:做一个写求职信的网站 (2024-01-15); CV Genius 求职信与简历调查 ; IO+:AI 改善了求职信,却没提高录用机会 (2026-07-30); Crown Staffing:求职信还重要吗 。核对日期:2026-11-12。

Fake links, made-up sources, clauses that aren't there: how to catch AI when it invents things

图片
On the OpenAI community forum, users have filed detailed reports of ChatGPT inventing web links and article titles, even in web-search mode and even after being asked to verify them. Another thread from May 2025 noticed one model adding hallucinated sources to the end of answers "very frequently". It isn't only casual users. In a Chinese developer forum, someone testing legal AI tools in June 2026 found one claiming a contract had a force-majeure clause in section 5. The contract had no such clause. Why AI makes things up A chatbot isn't looking facts up in a database. It predicts what text is most likely to come next. When the honest answer is "I don't know", a plausible-sounding guess often comes out instead, in the same confident tone as everything else. Names, numbers, citations and URLs are where this happens most. Three ways to get fewer made-up answers Give it the source. Don't ask it to remember an article or contract; paste it in a...

AI 推荐的频道不存在、合同条款是编的?3 个真实翻车案例和防坑办法

图片
AI 一本正经地胡编,有个专门的名字叫"幻觉"。我们在论坛里翻到几个很典型的真实例子: 推荐不存在的频道: 有人让 ChatGPT 推荐几个 YouTube 频道,事先还约定了"信息要可靠",结果好几个频道的名字、介绍、链接全是编的。(V2EX) 编出合同里没有的条款: 今年 6 月,有人测试法律 AI 审一份保密协议,AI 说第 5 条是不可抗力条款,可原文根本没有这一条;它还引用了一条对不上的法条。(V2EX) 编出假的网址和文章标题: OpenAI 官方论坛里,有用户反映 AI 在联网模式下照样会给出假链接、假标题,哪怕已经要求它核实。 为什么会这样 AI 不是在"查资料库",而是在猜下一句话最可能是什么。越冷门、越具体的东西(人名、书名、法条编号、网址、金额),越容易猜错,而且猜错时语气照样很肯定。 3 个让它少编的办法 把材料给它。 要它分析合同,就把合同原文贴进去,并要求"每个结论都标出在原文第几条"。 允许它说不知道。 提问时加一句:不确定的地方直接写"不确定",不要猜。 要出处就用会联网的工具 ,比如 Perplexity,然后把来源点开看一眼。 哪些内容一定要自己核对 数字、日期、金额、人名书名、链接、法律和医疗信息。尤其是要发给别人、要签字、要花钱的东西。 完整的核对清单: AI 一本正经地胡编怎么办 参考的真实讨论: V2EX:如何让 ChatGPT 的回答来自可靠的信息,而不是胡编乱造? (2023-03-07)、 V2EX:法律 AI 幻觉:你遇到过 AI 编造合同条款吗? (2026-06-16)、 V2EX:有没有发现,AI 的幻觉太严重了 (2026-08-04)、 OpenAI 社区:Hallucinated URLs + Fake Article Titles in Web Mode (2025-05-07)。核对日期:2026-11-12。

Using AI to read your lease? It's a great translator and a risky lawyer

图片
AI can genuinely help renters. One widely shared story: a tenant pasted his lease into ChatGPT after his landlord kept over $600 of his deposit, and it pointed him to a specific law the landlord had broken. He was careful to say it isn't a replacement for a lawyer. That caveat matters. When the Tenants' Union of NSW in Australia tested ChatGPT, Gemini and Copilot on tenancy questions, all three got things broadly right, but with mistakes. ChatGPT and Copilot both missed a document the landlord was legally required to provide. And a UK lettings company described a dispute where the tenant side cited three legal cases that didn't exist. What AI is good at Translating each clause into plain English Pulling out every cost: rent, deposit, fees, penalties, and when they're due Flagging unusual clauses so you know what to ask about What you still have to do Remove personal details first: names, ID numbers, bank details, full address. Make it point to the text. ...

租房合同看不懂?让 AI 逐条讲人话,但这 3 件事它替你做不了

图片
租房合同里的坑,V2EX 上讨论过很多次。有人租约到期后跟房东说下个月不租了,房东搬出合同里一句"到期自动续签";今年 1 月还有人发帖问:房东拿来的合同只写了租客信息、没写房东信息,要不要较真?——真有纠纷,连起诉谁都不知道。 在海外租房更难:一份英文租约十几页,满是专业词,看得人头大。 现在越来越多人让 AI 帮忙看合同。有用吗?有用,但要知道它的边界。 AI 能做好的 把每一条翻成大白话。 把"要花的钱、要担的责任、怎么提前退租"单独列出来。 标出看起来不常见、对你不利的条款,提醒你去问。 AI 会出错的地方(真实例子) 编出原文没有的条款: 今年 6 月有人测试法律 AI,它说保密协议第 5 条是不可抗力条款,原文根本没有。 漏掉当地法规的要求: 澳大利亚新南威尔士州租户联合会做过测试,ChatGPT 和 Copilot 都漏掉了一项房东必须提供的文件。 编出不存在的判例: 英国一家物业公司提到,有人在租赁纠纷里引用了三个判例,结果都不存在。 所以这样用最稳 先把姓名、证件号、银行账号、详细地址删掉或改成 XXX。 让 AI 逐条解释, 并要求每个结论写出对应原文第几条 ,原文没写的就说"原文没写"。 它列出的金额和日期,回原文逐条核对。 金额大、条款复杂、已经有纠纷的,找律师或当地租户组织。 可直接复制的逐条解释提示词: 用免费 AI 看懂英文合同和租约 参考的真实讨论: V2EX:合同写有到期自动续签,这样合法吗? (2015)、 V2EX:在外租房签合同,你会较真这点吗? (2026-01-16)、 V2EX:法律 AI 幻觉,你遇到过 AI 编造合同条款吗? (2026-06-16); Tenants' Union of NSW:AI 租房建议测试 ; CPS Homes:租客与 ChatGPT (2026-01-29)。核对日期:2026-11-12。

Hit Claude's free limit after a few messages? Here's how it's actually counted

图片
Some days you can chat with Claude for an hour on the free plan. Other days you're cut off after a handful of messages. It feels random. It isn't. How the free limit works It resets on a rolling five-hour window. When you hit it, Claude tells you the exact time you can continue. It isn't a message count. Usage depends on how much text each message carries. A long conversation, or a big file pasted in, uses up your allowance much faster than short questions in a fresh chat. Free accounts don't get a usage meter. Since August 2026, the usage page shows nothing to free users, so the reset time in the cutoff message is the number to trust. Make your free allowance last longer One task, one chat. When you finish something, start a new conversation instead of continuing the old one. Paste only what's needed. Send the three relevant pages, not the whole 80-page PDF. Say everything up front. One complete request costs less than five rounds of corrections...

Claude 额度"缩水"了?先搞懂它是怎么算的,再别信"拼车"

图片
"Claude 额度缩水了嘛?"——今年 4 月这个 V2EX 帖子下有十几条回复,有人说同样的活,现在一个 5 小时的额度撑不住了。1 月还有人发现额度好几天才刷新,才知道除了 5 小时,还多了一道"周额度",可是没找到详细说明。 免费用户的困惑更大:有时候聊几句就到上限,有时候能聊很久。 它到底怎么算 大约每 5 小时重置一次。 到上限时,Claude 会在提示里写明什么时候恢复。 不是按"几条消息"算,而是按对话的分量算。 对话越长、贴的文件越大,每条消息"吃"得越多。所以同样是 10 条,有时够、有时不够。 免费版看不到剩余用量的进度条,最可靠的仍然是那条提示里的恢复时间。 怎么让免费额度用得更久 一件事一个对话。 聊完开新的,别在一个对话里越聊越长。 文件只贴需要的部分。 整本资料塞进去,几条就满了。 到上限就换工具: ChatGPT、Gemini、Copilot 的免费版都能接着用。 千万别"拼车" 论坛里常见"额度合租""3 人车、5 人车"。同一个帖子里就有人说,两个人拼一个号被封了五个号。还有各种"$10 免费额度"的中转服务,要你把付款和账号交给第三方。省下的钱,可能连号一起赔进去。 到上限时的完整处理步骤: Claude 显示免费用量到上限怎么办 参考的真实讨论(V2EX): Claude 额度缩水了嘛? (2026-04-12)、 现在是怎么限制 claude 模型额度的? (2026-01-20)、 Claude Code 额度合租 (2026-01-27);用量规则参考 Claude Usage Limits (2026) 、 Claude 免费版用量说明 。核对日期:2026-11-12。

Too many AI tools? Stop asking "which is best" and ask "best for what"

图片
"Which AI is best?" is the wrong first question, and research backs that up. Section AI's 2026 report on AI proficiency found that most people's real barrier isn't prompting; it's not knowing what to use AI for. A Pew survey from February 2026 found 51% of US adults had never used a chatbot at all. So start with the task, then pick the tool. Every tool below has a free plan. Pick by what you're doing Writing, editing, everyday questions: ChatGPT, Claude or Gemini. For Chinese, DeepSeek, Kimi or Qwen. Research where you need sources: Perplexity. It lists where each answer came from. Reading long documents or your own notes: NotebookLM, Claude or Kimi. Making images: Adobe Firefly or Gemini. Turning text into speech: ElevenLabs. Making a song: Suno. The beginner's shortcut Get comfortable with one chat assistant first. Knowing one tool well beats knowing five a little. Bookmark a backup. When the first one is busy or out of free ...

AI 那么多,新手到底该用哪个?别问"哪个最强",先问"我要干嘛"

图片
"用 AI 写小说,哪个模型比较合适?""目前哪个 AI 软件写代码最强?""免费 ChatGPT 和免费 Gemini 哪个好?"——在 V2EX 上,这类"哪个最好"的问题隔几天就有一个。 看下面的回复会发现: 答案几乎都是"看你要干什么"。 写小说的帖子里,有人推荐 Kimi,也提醒太长了会"吃设定";有人说记忆能力是大问题。写代码的帖子里,几乎一边倒推荐某几个专门工具。拿同一个问题考两个免费 AI 的帖子里,有人觉得一个"更好相处",一个"更专业"。 按要做的事挑(都有免费版) 写东西、改文字、日常问答: 海外用 ChatGPT、Claude、Gemini;中文用 DeepSeek、Kimi、通义千问、豆包。 查资料、要看出处: Perplexity,回答后面会列来源。 读很长的文件、整理资料: Gemini Notebook(NotebookLM)、Kimi、Claude。 画图: Adobe Firefly、Gemini。 文字转语音: ElevenLabs。 做一首歌: Suno。 新手最省事的做法 先只用一个聊天类 AI,把它用熟。 工具换来换去,不如把一个用明白。 再收藏一个备用的。 第一个卡住(繁忙、额度用完)时直接换。 真有专门的需求 (画图、配音、读整本书)再去试专门工具。 "最强"每个月都在变,"适合你手上这件事"才是你真正要找的。 按用途分好类、附官方入口的选择页: AI 太多了,不知道该用哪个怎么办 参考的真实讨论(V2EX): 用 ai 写小说的话,哪个模型比较合适 (2026-02-17)、 目前哪个 ai 软件写代码最强? (2025-03-11)、 免费 ChatGpt 和免费 Gemini 的对比 (2026-01-24)。核对日期:2026-11-12。

Free ChatGPT suddenly got dumber? You may have run out of the good model

图片
On the official OpenAI community forum, one free user said they got about seven messages in before ChatGPT told them they'd hit the limit and would have to wait five hours. Another said the reset timer showed the wrong time for days. In October 2026, someone posted a feature request for something simple: let free users see how much they have left. If ChatGPT has felt noticeably worse halfway through a chat, there's a likely reason. What's actually happening The free plan gives you a limited number of messages on the best model within a time window. When you use them up, ChatGPT often doesn't stop. It quietly switches you to a smaller model until the window resets. The notice is easy to miss, so it just feels like it got dumber. The exact number changes, and articles online disagree. The only figure you can rely on is the reset time ChatGPT shows you in its own message. What to do Check the reset time in the notice. Small tasks like fixing a sentence or trans...

免费版 ChatGPT 突然变笨了?可能是你的"好模型次数"用完了

图片
今年 1 月,V2EX 上有人拿同一个问题去问免费 ChatGPT 和免费 Gemini,结果 ChatGPT 答得一塌糊涂。他的疑问很直接:是因为我不是 Plus 吗? 英文社区里也有类似的抱怨。有人在 OpenAI 官方论坛说,自己晚上才打开 App,只发了 7 条左右就提示免费额度用完、5 小时后恢复。还有人说倒计时显示的恢复时间不对,等了好几天还在跳。今年 10 月,又有人在论坛提建议:希望免费用户能看到自己还剩多少额度。 这些抱怨背后,其实是同一件事。 为什么会"突然变笨" 免费版 ChatGPT 的好模型,在一段时间内只能用有限的次数。用完以后,它不一定会停下来,而是常常 自动换成一个小一点的模型 继续回答。界面上的提示很容易被忽略,你只会觉得"它怎么变笨了"。 次数到底是多少,官方会根据情况调整,网上各种说法对不上。最靠谱的只有一个: ChatGPT 弹出的提示里写的恢复时间。 遇到了怎么办 看提示里的恢复时间。简单的事(改句子、翻译一段)小模型也够用,复杂的事等到点再问。 急用就换一个免费 AI:Claude、Gemini、Copilot 都有免费版,中文还可以用 DeepSeek、Kimi。 省着用:一次把要求说清楚,比来回改十次省得多。 别碰的东西 网上一直有人分享"免费账号""共享账号""低价代充"。这类账号来路不明,被盗、被封的风险都在你身上。想升级,只在官方 App 或官网里付费。 完整的急救步骤和各工具官方入口: ChatGPT 提示免费额度用完了怎么办 参考的真实讨论: V2EX:免费 ChatGpt 和免费 Gemini 的对比 (2026-01-24);OpenAI 社区: I need help with the message limit (2025-04-01)、 My timer resets at the wrong times (2025-03-06)、 Separate Usage Tracker and Early Limit Notifications (2026-10-30); It's FOSS 社区:免费版到上限后换用其他模型的提示 。核对日期:2026-11-12...

Can recruiters tell your resume was written by AI? What the 2026 surveys actually say

图片
Using AI on your resume is now normal. In Resume Genius's 2026 survey of 1,000 US hiring managers, 77% said many of the resumes they see look partly or fully AI-written. The same survey found that 80% of hiring managers believe they can tell. But when TopResume actually tested 600 hiring managers, only about one in three picked the AI-written resume out of four, barely better than guessing. So the real question isn't "will they catch me?" It's "does my resume sound generic?" Because that is what hiring managers say gives it away. What makes a resume "feel" AI-written When Resume Genius asked hiring managers what tipped them off, the top answers were: Unnatural phrasing or tone Repetitive or overly generic language Vague or inflated descriptions Notice what's not on the list: "used AI at all." In TopResume's survey, about half of hiring managers said using AI for drafting or proofreading is fine. What they push ba...

用 AI 写的简历,HR 一眼就看得出来?3 个办法让它"有人味儿"

图片
中国青年报今年春天调查了 2040 名 2026 届毕业生,超过 86% 的人在春招里用过 AI 工具,改简历是最常见的用法之一。 用的人多了,问题也来了。知乎上一位做了多年 HR 的网友说得很直接:AI 润色的简历句式工整、话术专业,却少了最关键的"灵魂"。V2EX 上也有人试着让 AI 根据自己的项目写经历,结论是"效果不是很满意,没有人味儿"。 还有一种担心更实际:你只给了三行字,AI 还你一段连自己都认不出来的辉煌履历。面试一问,答不上来。 所以问题不是"能不能用 AI 写简历",而是 怎么用,才能又快又像你自己。 办法一:先给素材,别让它"凭空写" AI 写得空,多半是因为你给得少。动手前,先花 5 分钟把每段经历写成这样一句话: 当时是什么情况、你做了什么、结果怎么样。 有数字就写数字,没有就写具体的事。 办法二:明确告诉它"不许编" 在提示词里直接写上:只用我给的信息,不要编造经历、数字或证书;不清楚的地方写"待补充"。这一句能挡住大部分"创造性发挥"。 办法三:最后一遍,用你自己的话改 AI 出完稿,逐条读,问自己两个问题: 这件事是真的吗?面试官追问,我讲得出细节吗? 讲不出来的,换成你能讲清楚的说法。HR 想看的是你,不是一套标准话术。 照着做:10 分钟、0 元 完整步骤和可以一键复制的提示词(已经写好"不许编"那一句),都在这一页: 用免费 AI 写一份简历 。要投具体岗位,还可以接着看 用免费 AI 写求职信 。 本文参考: 中国青年报《你的AI求职"搭子"靠谱吗?》 (2026-05-12,含知乎网友回答)、 V2EX:有大佬使用过 AI 来阅读项目代码来生成简历上的项目经验吗? (2026-02-26)、 CSDN:担心 AI 会瞎编经历,怎么用 AI 只优化不造假? 。核对日期:2026-11-12。

DeepSeek says "server is busy"? Don't install a "fix" yet. Do this instead

图片
"The server is busy. Please try again later." If you use DeepSeek, you've probably seen it. It isn't only you: full DeepSeek outages have made the front page of Hacker News more than once, most recently in September 2026, when the web app, mobile app and API all went down together. Search for a fix and you'll land on pages listing ten or more "solutions". Most of them are written for developers. Here's the beginner version. What the message means Too many people are asking DeepSeek things at the same moment, so it turns some requests away to protect itself. Your computer is fine, your internet is fine, and your account hasn't been banned. The "fixes" beginners should skip Browser scripts and unknown extensions that promise to auto-retry for you. They get access to your browser, and you don't know who wrote them. Third-party routes that need an API key. Some guides point you to other platforms that run the same model. The...

DeepSeek 又"服务器繁忙"?先别急着找"加速器",看完这篇再说

图片
"今天 DeepSeek 崩了吗?"——这句话在 V2EX 上隔一阵就有人发一次。今年 5 月有人问过,11 月 3 日又有人发帖,正文就一句:服务器繁忙,请稍后重试。 还有人说得更具体:每天第一次问答能用,第二次就开始繁忙。 如果你也正盯着这行字发愁,这篇只讲三件事: 它到底是什么意思、网上流传的"解决办法"哪些先别碰、你现在能做什么。 一、这不是你的问题 "服务器繁忙"的意思是:同一时间用的人太多,DeepSeek 先挡住一部分请求,免得整个系统崩掉。不是你的电脑坏了,不是网络不好,也不是账号被封。 想知道是不是"大家都打不开",最简单的办法:搜一下"DeepSeek 崩了"。如果刚刚有很多人在问,那就是全网的问题,你能做的只有等,或者先换一个 AI。 二、网上那些"解决办法",小白先别碰 我们翻了论坛里的讨论,常见的"办法"大概三类: 浏览器脚本、插件: 有人做了脚本帮你自动重试,作者自己都把它叫作"饮鸩止渴版"。装来路不明的脚本和插件,等于把浏览器权限交给陌生人。 第三方 App、"满血版"平台: 很多要你去别处申请 API Key、充值、填一串网址做配置。对新手来说步骤多、容易出错,还可能花冤枉钱。 "DeepSeek 会员""加速器": 官方网页版和 App 本来就免费,花钱买这类东西,大概率是被割韭菜。 这些办法不是都不能用,但它们是给爱折腾的人准备的。如果你只是想把手上这件事做完,下面的办法更快。 三、你现在能做的 3 步 先把问题复制下来 ,免得刷新后丢了。 等一两分钟再发 ,别连着狂点。 急用就换一个免费 AI :把问题粘到 Kimi、通义千问、豆包;在海外可以用 Gemini、ChatGPT、Claude 的免费版。 一页看完的急救步骤,附各个工具的官方入口: DeepSeek 显示"服务器繁忙"怎么办 一句话总结 重要的事,别只押在一个 AI 上。平时多收藏一个能用的,卡住时直接换,比研究任何"破解办法"都省事。 本文参考的真实讨...

Free AI access: check the limits before you start

图片
  Before using a free offer, ask what is included, which accounts qualify, and what happens when the allowance runs out. Three common types Free plans:  access through a website or app, with possible usage or feature limits. API credits:  an allowance for software that calls a model; this may be separate from chat access. Promotional rewards:  credits that may require a check-in, a new account, or other steps. Before claiming Follow the provider’s official website. Check region, account eligibility, payment requirements, expiry, and any charges after the allowance is used. An old number on a third-party page is not proof that an offer is still available. After claiming Check the actual balance or transaction history in your provider account. A manually recorded card on AIHOT is a personal record, not confirmation that a provider issued credits. It does not perform a check-in for you. Try one useful task Draft a short invitation or organize a note. Free offers can cha...

Your first AI task: write a clear invitation

图片
A good first AI task is something you can check yourself. Try drafting a short invitation for an imaginary book club. You do not need real names or contact details. 1. Give the AI clear facts Open your chosen AI chat tool and try this prompt: Write a friendly invitation for people attending a book club for the first time. It is this Saturday at 3 pm in the community room and lasts one hour. Ask everyone to bring a book they like. Keep it under 80 words. Do not invent details. Ask me about anything that is unclear. 2. Ask for one improvement If it sounds too formal, say: “Make it sound like a message to a friend. Keep the time, place, and what to bring.” If it is too long, ask for three short lines. 3. Check before sharing Are the time, place, and duration correct? Did the AI invent a fee, contact person, or other arrangements? Will readers know what to do next? Replace “this Saturday” with a verified date before sending. Replace the imaginary venue and arrangements with real ones, too....

第一次用 AI:从一条清楚的通知开始

图片
第一次使用 AI,可以先完成一件能自己检查的小事:写一条通知。这里用虚构的读书会做练习,不需要提供真实姓名或联系方式。 1. 给它明确的材料 打开你选择的 AI 工具,在对话框中输入下面这段话。 请帮我写一条读书会通知,发给第一次参加的朋友。时间是本周六下午 3 点,地点是社区活动室,活动持续 1 小时,请大家带一本喜欢的书。语气友好,100 字以内。不要添加我没有提供的信息;不确定的地方请问我。 2. 看完,再补一句要求 如果太正式,就说:“请更像朋友聊天,保留时间、地点和携带物品。”如果太长,就说:“压缩成三行,每行只讲一件事。”这就是一次有效的追问。 3. 发送前检查三件事 时间、地点和活动时长有没有写错? 有没有擅自添加报名费用、联系人或其他安排? 读者能不能一眼看出自己要做什么? “本周六”在转发后可能有歧义,正式使用时请换成核实过的具体日期。练习中的地点和活动安排也要替换成真实信息。 今天学会的一个方法 告诉 AI:我要做什么、给谁看、有哪些已知信息、希望输出多长。然后修改并检查。先把一件小事做好,再尝试更复杂的任务。 想跟着做一遍?这节课的完整练习和可直接复制的提示词在网站上: 用 AI 写一份活动通知(新手第 1 课)

免费 AI 怎么用:先看懂额度,再开始

图片
看到“免费”时,先问三个问题:免费的是哪个功能?适用于什么账号?用完后会发生什么? 常见的三种形式 免费版本: 在网页或客户端中使用,可能有次数、速度或功能限制。 API 额度: 供程序调用模型,通常与网页聊天的额度分开。 活动积分: 可能要求签到、首次注册或完成指定步骤,是否有效以官方活动说明为准。 领取前核对 确认入口来自厂商官网;检查地区、账号资格、是否需要绑定付款方式、有效期和超额费用。不要仅凭第三方页面上的旧数字判断今天还能领多少。 领取后核对 到厂商账号中查看实际余额或领取记录。本站卡册中的“已记录”是你的手动记录,不代表厂商已发放额度,也不会代你签到。 先完成一个小任务 选一个工具试写一条通知或整理一段笔记。免费规则会变动,真正值得保留的是你学会的使用方法。 16 个常用 AI 工具的免费条件和官方入口,我们整理在一页里,规则没核实的照实标注: AI 免费额度一览