Workbuddy 接入 Hy3 限时免费:不知道用来做什么的,可以看我这篇Workbuddy Connects with Hy3 for Limited-Time Free Trial: Uncertain What to Use It For? Check Out My Article


AI free token/AI免费积分
图片

 

Hi,大家好!我是蜡笔,你的AI入门学习指南💡

 

 

今天主要来分享下Workbuddy已接入Hy3,而且限时免费(7.6-7.22),以及Hy3小小实测分享,和给你们一些要怎么去用它的思路。

 

腾讯混元发布 Hy3,官方的说明总结:

 

腾讯混元发布 Hy3,在推理、智能体、长上下文等任务上显著进步,比肩 2~5 倍参数规模旗舰模型。内部盲测均分 2.67/4,优于 GLM5.1(2.51/4)。幻觉率从 12.5% 降至 5.4%,常识错误率从 25.4% 降至 12.7%,多轮问题率降至 7.9%,MRCR 从 42.9% 升至 75.1%。任务解决率从 72% 跃升至 90%,平均耗时缩短 34%。API 价格为输入 1 元、输出 4 元(每百万 tokens),命中缓存 0.25 元。已以 Apache 2.0 协议在 GitHub、HuggingFace 等平台开源。

 

个人实操感觉:比之前的 Hy3 Preview 相比,这是一次大版本的升级。

 

Image

 

很多人会问,虽然免费,但我要用hy3来做什么?

 

免费的东西总是让人想立刻打开试试,但真点进去,又容易卡在第一步:到底该从哪里开始?

然后我就发现了一个挺实用的入口:wb 左侧栏目里的【更多 - 灵感】,已经整理了很多常见工作流,而且都沉淀成了可复用的任务起点。

 

图片

 

无论是办公协同、内容创作,还是商业运营、旅行出行都能找到对应案例。每个案例都可以直接参考,也能在原有基础上改成自己的需求。

对刚开始使用的人来说,它最大的价值,其实就是帮你少一点空白页焦虑,先从一个现成思路开始,再慢慢改成自己的工作流。

 

Image

 

随机点开一个,可以先浏览下内容和样式,如果觉得符合自己某个需求,直接点【做同款】,或者先💗收藏起来。

 

Image

 

点击做同款,即回到对话页面;会显示那份内容的提示词和对应引用的专家团队或者技能(Skill),这里要注意⚠️:除非你想做完全一样的报告,否则提示词,还是要修改下再去执行。

 

Image

 

用任意的ai chat去修改下提示词,让ai根据你的主题“”去修改那段参考的提示词结构;案例如下:

 

Image

 

我改好的提示词,也可以参考下:

 

  
帮我做一份《经济学人》/FT 财经社论报刊风格的单文件 HTML 报告,主题为《AI 对中国就业市场的冲击:一场岗位结构的再定价》。
整体要求:
采用米白浅暖底、衬线大标题、暗红主色点缀,多栏报刊排版、首字下沉、财经社论式视觉气质。版式参考《经济学人》深度专题与 FT 周末长文风格,强调冷静、克制、专业、略带判断力。全部 CSS/JS 内联,无外链,无 emoji,简体中文,响应式适配桌面与移动端。
内容结构要求:
一、刊名报头与导语摘要
设计一个类似财经报刊专题页的刊名报头,例如:
THE EMPLOYMENT BRIEF
AI 与中国就业市场年度观察
导语摘要需用社论语气概括核心判断:
AI 对中国就业市场的影响,并不是简单地让工作消失,而是让可标准化、可流程化、可文本化、可代码化的岗位价值被重新定价。中国就业总量仍保持韧性,但岗位结构正在分化:一边是制造业、服务业和基层岗位仍有用工缺口,另一边是白领岗位、内容岗位、初级研发、数据处理、客服、运营、设计助理等岗位面临效率工具带来的替代与压缩。
二、核心指标条
用国家统计局公报真实数据做核心指标条,至少包含:
城镇新增就业:1267 万人
全国城镇调查失业率均值:5.2%
全国就业人员:72504 万人
城镇就业人员占比:65.6%
指标条下方标注:
数据来源:国家统计局《2025 年国民经济和社会发展统计公报》
以上为就业市场宏观基准数据,用于说明 AI 冲击发生在一个总量就业相对稳定、结构调整持续深化的背景中。
三、社论正文:两条主线
正文采用两栏或三栏报刊排版,写成财经社论风格,不要写成普通科普文。
主线一:总量稳,冲击被结构吸收
说明中国就业市场并未因为 AI 出现总量崩塌。城镇新增就业仍完成较高规模,调查失业率均值保持在 5.2%。但这并不意味着 AI 影响不大,而是说明 AI 的影响首先表现为岗位内容变化、招聘门槛变化、薪酬结构变化和入门岗位减少,而不是立刻体现为失业率剧烈上升。
重点分析:
AI 先改变企业的人效模型,而不是马上替代全部岗位。
企业更倾向于少招初级岗位、压缩外包、提升单人产出。
就业统计中的稳定,可能掩盖岗位质量、职业入口和技能溢价的变化。
AI 冲击更像慢变量,不像金融危机式的断崖冲击。
主线二:结构分,岗位被重新分层
从岗位类型角度分析 AI 的影响,至少分为四类:
高风险岗位:客服、基础文案、初级设计、基础数据处理、简单代码、报表制作、资料整理、翻译润色、低复杂度运营。
中风险岗位:财务分析、人力资源、法务初稿、市场策划、产品助理、项目文档、测试、实施顾问、售前方案等。
低风险岗位:复杂制造现场、客户关系、组织协调、线下服务、非标项目管理、复杂决策岗位。
受益岗位:AI 产品经理、智能体搭建、数据治理、行业解决方案、AI 应用运营、AI 培训、自动化流程设计、AI 内容导演、企业数字化转型岗位。
强调一个核心判断:
AI 不会平均地冲击所有人,它会优先挤压低经验、低壁垒、重复性强、交付物标准化的岗位,同时放大具备行业理解、流程设计、审美判断、复杂沟通和系统集成能力的人。
四、内联 SVG 图表
请设计 2-3 个内联 SVG 图表,风格要像报刊信息图,不要花哨。
图表 1:月度城镇调查失业率折线图
要求包含 5% 参考线,展示全年失业率围绕 5%左右波动。若缺少完整官方月度数据,可明确标注为示意测算或基于公开月度数据整理。
图表 2:全龄失业率 vs 青年失业率柱状对比
用柱状图表达青年就业压力高于全龄总体就业压力。注意标注青年失业率口径可能涉及不含在校生等统计口径变化,不要偷懒装成绝对真理,毕竟数据口径这种东西,人类社会最擅长把简单问题变复杂。
图表 3:AI 替代风险岗位分布图
用横向条形图或风险矩阵展示不同岗位类型受 AI 影响程度,例如:
客服与基础文案:高
初级设计与简单视频剪辑:高
基础数据处理与报表:高
初级代码与测试:中高
财务、人力、法务初稿:中
制造现场、客户关系、复杂项目管理:低到中
AI 应用、数据治理、行业解决方案:受益
所有图表必须使用内联 SVG,不依赖外部库。
五、结构性议题卡片
设计 4-6 张议题卡片,每张有标题、短观点和影响说明。
建议卡片包括:
青年就业入口收窄
AI 让企业减少实习生、助理、初级岗位,把原本用于培养新人的任务交给工具完成。问题不只是少了岗位,更是少了练级场。
白领岗位的隐性压缩
许多企业不会直接宣布 AI 裁员,而是通过冻结编制、提高 KPI、减少外包、合并岗位来完成效率再分配。
招工难与求职难并存
制造业一线、蓝领技能、养老护理、线下服务仍存在用工缺口,但高校毕业生更集中流向办公室岗位,供需错配被进一步放大。
慢就业与考公考研升温
部分年轻人面对不确定性选择延迟进入就业市场,继续升学、考编、考证或低强度过渡,就业压力被时间推迟。
AI 技能变成新门槛
会不会用 AI,正在从加分项变成默认项。真正稀缺的不是会输入提示词,而是能把 AI 嵌入业务流程、形成稳定交付的人。
企业数字化岗位再升值
懂业务、懂流程、懂数据、懂系统集成的人会更有优势。AI 不是孤立工具,它最终需要落到企业流程、权限、数据、审批、风控和组织协同里。
六、专题分析:对中国就业市场的五个具体冲击
请单独设置一个深度分析区,标题为:
AI 冲击的五个就业后果
内容包括:
岗位数量:部分初级岗位减少,但新岗位也在产生。
岗位内容:从执行型任务转向判断型、协调型、整合型任务。
招聘门槛:企业要求候选人同时具备专业能力和 AI 工具使用能力。
薪酬结构:普通执行岗位议价能力下降,复合型人才溢价上升。
职业路径:过去从简单任务入门、逐步积累经验的路径被削弱,新人更难获得低风险练手机会。
七、制造业与企业信息化视角
增加一个面向制造企业的分析区,标题为:
制造企业中的 AI:不是替代工人,而是重写白领流程
请结合制造企业实际场景分析 AI 对岗位的影响:
采购:供应商资料整理、询价比价、合同初稿、风险提示可被 AI 辅助。
财务:费用审核、凭证摘要、报表解释、预算分析可被 AI 提效。
质量:异常报告、8D 报告、质量追溯分析可被 AI 辅助生成。
生产计划:计划解释、异常预警、库存联动分析可被 AI 支持。
IT/信息化:需求文档、测试用例、用户手册、系统配置说明、SQL 初稿、报表逻辑整理会被大幅提效。
管理层:AI 会让经营分析更及时,但前提是企业数据治理要跟上。
核心判断:
在制造企业里,AI 最先改变的不是产线工人,而是办公室里大量依赖文档、表格、流程、系统和沟通的岗位。真正的竞争力来自业务流程数字化、数据标准化和组织协同能力。
八、结论区:社论式判断
用一段有力量的总结收尾:
AI 对中国就业市场的冲击,不是一个简单的失业故事,而是一场岗位价值的重新排序。总量就业仍能维持稳定,但结构性分化会越来越明显。未来更危险的,不是某个工具突然取代某个人,而是一个人长期停留在只会执行、不会判断、只会使用工具、不会设计流程的位置上。就业市场不会消失,但它会变得更挑剔。
九、页脚与数据说明
页脚必须明确列出:
数据来源机构:国家统计局、人力资源和社会保障部等。
报告名与年份:如《2025 年国民经济和社会发展统计公报》。
哪些数据为官方真实数据。
哪些图表或岗位风险判断为示意测算、基于公开资料与行业观察整理。
明确声明:本报告为财经社论风格的信息可视化页面,不构成正式统计报告或投资建议。
十、代码要求
输出一个完整 HTML 文件。
CSS、JS 全部内联。
不使用外部字体、图片、图标、脚本或 CDN。
使用语义化 HTML。
页面需具备良好的响应式表现。
图表使用内联 SVG。
色彩控制在米白、深灰、暗红、浅金、淡棕范围内。
排版要像高级财经报刊专题,而不是普通网页模板。
不要使用 emoji。
不要输出解释,只输出完整 HTML 代码。

 

这是我执行生成的结果:放一些局部截图,样式排版完全符合我需求,稳定性保持很好;得到的报告对我也有一定意义,具体的数据看数据源是科学的,我没有完全去验证。

 

Image

 

Image

 

Image

 

 

同样的操作,我还做了一个读书笔记网页,刚好今天读到卡兹克的文章;就用《人生设计课》来做,非常美丽,很喜欢。

 

Image

 

Image

 

Image
Image

 

Image
Image

 

Image
Image

 

Image
Image

 

Image
Image

 

 

H

Image

 

 

H

 

Hy3也支持图片生成:

 

至于效果,ennnnnm,你们自评。

 

Image

 

 

当然啦,你还可以安装Skill以及使用看看。我今天由于使用过于频繁,已被提醒限额了。

 

图片

Hi everyone! I'm Labi, your AI beginner's guide 💡

Today I'm mainly sharing that Workbuddy has integrated Hy3, and it's free for a limited time (7.6–7.22), along with a small hands-on review of Hy3, and some ideas for how you might use it.

Tencent Hunyuan released Hy3, and the official description is summarized as:

Tencent Hunyuan released Hy3, showing significant progress on reasoning, agents, long-context and other tasks, comparable to flagship models with 2–5x its parameter size. In internal blind testing, it scored 2.67/4, outperforming GLM5.1 (2.51/4). The hallucination rate dropped from 12.5% to 5.4%, the common-sense error rate dropped from 25.4% to 12.7%, the multi-turn question rate dropped to 7.9%, and MRCR rose from 42.9% to 75.1%. Task resolution rate jumped from 72% to 90%, with average time reduced by 34%. API pricing is 1 yuan for input and 4 yuan for output (per million tokens), with 0.25 yuan for cache hits. It has been open-sourced under the Apache 2.0 license on platforms such as GitHub and HuggingFace.

My personal hands-on feeling: compared to the previous Hy3 Preview, this is a major version upgrade.

Image

Many people will ask: even though it's free, what should I use Hy3 for?

Free things always make you want to open and try them right away, but once you actually click in, it's easy to get stuck at the first step: where exactly do I start?

Then I found a pretty practical entry point: in wb's left sidebar, under [More - Inspiration], many common workflows have already been organized, and they've all been distilled into reusable task starting points.

图片

Whether it's office collaboration, content creation, business operations, or travel, you can find corresponding examples. Each example can be referenced directly, and you can also adapt it to your own needs based on the original.

For people just starting out, its biggest value is really helping you reduce blank-page anxiety — start from an existing idea, then slowly turn it into your own workflow.

Image

Click on any one at random, and you can first browse the content and style. If you feel it matches one of your needs, just click [Make Same Style], or first 💗 favorite it.

Image

Clicking "Make Same Style" returns you to the chat page; it will show that content's prompt and the corresponding referenced expert team or Skill. ⚠️ Note here: unless you want to make an exactly identical report, you should still modify the prompt before executing it.

Image

Use any AI chat to modify the prompt, and let the AI modify the reference prompt structure according to your topic; example as follows:

Image

You can also refer to my revised prompt:

Create a single-file HTML report in the style of The Economist / FT financial editorial newspaper, themed "AI's Impact on China's Job Market: A Re-Pricing of Job Structures". Overall requirements: Use a warm off-white background, serif large headlines, dark red accent color, multi-column newspaper layout, drop caps, and a financial editorial visual style. Layout reference: The Economist deep-dive features and FT weekend long-read style, emphasizing calm, restrained, professional, and slightly judgmental. All CSS/JS inline, no external links, no emoji, Simplified Chinese, responsive for desktop and mobile. Content structure requirements: 1. Publication masthead and lede summary. Design a masthead similar to a financial newspaper special feature page, for example: THE EMPLOYMENT BRIEF — AI and China's Job Market Annual Review. The lede summary must use an editorial tone to summarize the core judgment: AI's impact on China's job market is not simply making jobs disappear, but re-pricing the value of jobs that can be standardized, process-driven, text-based, and codified. China's total employment remains resilient, but job structure is diverging: on one side, manufacturing, services, and grassroots positions still have labor gaps; on the other side, white-collar roles, content roles, junior R&D, data processing, customer service, operations, and design assistant roles face substitution and compression from efficiency tools. 2. Core indicator bar. Use real data from the National Bureau of Statistics communiqué for a core indicator bar, including at least: Urban new employment: 12.67 million people; National urban surveyed unemployment rate average: 5.2%; National employed persons: 725.04 million; Urban employed persons share: 65.6%. Below the indicator bar, note: Data source: National Bureau of Statistics, "Statistical Communiqué of the People's Republic of China on the 2025 National Economic and Social Development". These are macro baseline data for the job market, used to show that the AI impact occurs against a backdrop of relatively stable total employment and continuously deepening structural adjustment. 3. Editorial body: two main threads. Use two- or three-column newspaper layout, written in financial editorial style, not as an ordinary popular-science article. Thread 1: Total volume is stable; the impact is absorbed by structure. Explain that China's job market has not collapsed in total volume because of AI. Urban new employment still reaches a relatively high scale, and the surveyed unemployment rate average stays at 5.2%. But this does not mean AI's impact is small; rather, it means AI's impact first shows up as changes in job content, hiring thresholds, pay structures, and fewer entry-level positions, not immediately as a sharp rise in unemployment. Key analysis: AI first changes companies' people-efficiency models rather than immediately replacing all jobs. Companies tend to hire fewer junior roles, cut outsourcing, and raise output per person. Stability in employment statistics may mask changes in job quality, career entry points, and skill premiums. AI's impact is more like a slow variable than a cliff-like financial-crisis shock. Thread 2: Structure is diverging; jobs are being re-layered. Analyze AI's impact by job type, at least four categories: High-risk jobs: customer service, basic copywriting, junior design, basic data processing, simple coding, report making, document sorting, translation and polishing, low-complexity operations. Medium-risk jobs: financial analysis, HR, legal first drafts, marketing planning, product assistant, project documentation, testing, implementation consultant, pre-sales solutions, etc. Low-risk jobs: complex manufacturing sites, client relationships, organizational coordination, offline services, non-standard project management, complex decision-making roles. Benefiting jobs: AI product managers, agent building, data governance, industry solutions, AI application operations, AI training, automation process design, AI content directors, enterprise digital transformation roles. Emphasize one core judgment: AI will not hit everyone equally; it will first squeeze jobs that are low-experience, low-barrier, repetitive, and with standardized deliverables, while amplifying people who have industry understanding, process design, aesthetic judgment, complex communication, and system integration capabilities. 4. Inline SVG charts. Design 2–3 inline SVG charts, styled like newspaper infographics, not flashy. Chart 1: Monthly urban surveyed unemployment rate line chart. Include a 5% reference line, showing the annual unemployment rate fluctuating around 5%. If complete official monthly data is unavailable, clearly mark it as an illustrative estimate or based on public monthly data compiled. Chart 2: All-age unemployment rate vs youth unemployment rate bar comparison. Use a bar chart to show that youth employment pressure is higher than overall all-age employment pressure. Note that the youth unemployment rate definition may involve statistical-caliber changes such as excluding in-school students; don't lazily present it as absolute truth, because data calibers are exactly the kind of thing human society is best at making complicated. Chart 3: AI substitution risk job distribution chart. Use horizontal bars or a risk matrix to show how different job types are affected by AI, for example: Customer service and basic copywriting: high; Junior design and simple video editing: high; Basic data processing and reports: high; Junior coding and testing: medium-high; Finance, HR, legal first drafts: medium; Manufacturing sites, client relations, complex project management: low to medium; AI applications, data governance, industry solutions: benefiting. All charts must use inline SVG and not depend on external libraries. 5. Structural issue cards. Design 4–6 issue cards, each with a title, short viewpoint, and impact description. Suggested cards include: Youth employment entry narrowing — AI makes companies reduce interns, assistants, and junior roles, handing tasks originally meant for developing newcomers to tools. The problem is not just fewer jobs, but fewer training grounds. Hidden compression of white-collar roles — Many companies will not directly announce AI layoffs, but instead complete efficiency redistribution through hiring freezes, higher KPIs, reduced outsourcing, and merged roles. Labor shortages and job-hunting difficulties coexist — Manufacturing frontline, blue-collar skills, elderly care, and offline services still have labor gaps, but college graduates flow more heavily toward office jobs, and the supply-demand mismatch is further amplified. Slow employment and rising civil-service/postgraduate exam fever — Some young people choose to delay entering the job market amid uncertainty, continuing into further study, exam prep, certification, or low-intensity transitions, pushing employment pressure into the future. AI skills becoming a new threshold — Whether you can use AI is shifting from a plus item to a default requirement. What is truly scarce is not being able to type prompts, but being able to embed AI into business processes and form stable deliverables. Enterprise digital roles appreciating again — People who understand business, processes, data, and system integration will have more advantage. AI is not an isolated tool; it ultimately must land in enterprise processes, permissions, data, approvals, risk control, and organizational coordination. 6. Special analysis: five specific impacts on China's job market. Set up a separate deep-analysis section titled: Five employment consequences of the AI shock. Content includes: Job quantity: some junior roles decrease, but new roles are also being created. Job content: shifting from execution-type tasks to judgment-type, coordination-type, and integration-type tasks. Hiring thresholds: companies require candidates to have both professional ability and AI tool proficiency. Pay structure: bargaining power for ordinary execution roles declines, while the premium for compound talent rises. Career paths: the old path of entering through simple tasks and gradually accumulating experience is weakened, and newcomers find it harder to get low-risk practice opportunities. 7. Manufacturing and enterprise informatization perspective. Add an analysis section aimed at manufacturing enterprises, titled: AI in manufacturing enterprises: not replacing workers, but rewriting white-collar processes. Analyze AI's impact on jobs using real manufacturing scenarios: Procurement: supplier data organization, inquiry and price comparison, contract drafts, and risk alerts can be AI-assisted. Finance: expense review, voucher summaries, report explanations, and budget analysis can be made more efficient by AI. Quality: anomaly reports, 8D reports, and quality traceability analysis can be AI-assisted. Production planning: plan explanations, anomaly alerts, and inventory linkage analysis can be supported by AI. IT/informatization: requirement documents, test cases, user manuals, system configuration notes, SQL drafts, and report logic organization will be greatly streamlined. Management: AI will make business analysis more timely, but only if enterprise data governance keeps up. Core judgment: In manufacturing enterprises, AI first changes not production-line workers, but the many office roles that rely heavily on documents, spreadsheets, processes, systems, and communication. Real competitiveness comes from business process digitization, data standardization, and organizational coordination capability. 8. Conclusion: editorial-style judgment. End with a powerful summary: AI's impact on China's job market is not a simple unemployment story, but a reordering of job value. Total employment can remain stable, but structural divergence will become increasingly obvious. In the future, the greater danger is not a tool suddenly replacing someone, but a person remaining for a long time in a position that can only execute, cannot judge, can only use tools, and cannot design processes. The job market will not disappear, but it will become pickier. 9. Footer and data notes. The footer must clearly list: Data source institutions: National Bureau of Statistics, Ministry of Human Resources and Social Security, etc. Report name and year: e.g., "Statistical Communiqué on the 2025 National Economic and Social Development". Which data are officially real data. Which charts or job-risk judgments are illustrative estimates, compiled from public materials and industry observations. Explicit disclaimer: This report is an information visualization page in financial editorial style and does not constitute a formal statistical report or investment advice. 10. Code requirements. Output one complete HTML file. All CSS and JS inline. No external fonts, images, icons, scripts, or CDNs. Use semantic HTML. The page must have good responsive behavior. Charts use inline SVG. Colors controlled within off-white, dark gray, dark red, light gold, and light brown. Layout should look like a premium financial newspaper special feature, not an ordinary web template. No emoji. No explanation output; only complete HTML code.

This is the result I generated from execution: here are some partial screenshots. The style and layout fully matched my needs, and stability stayed very good; the report I got also has some meaning to me. The specific data looks scientific based on the data sources, though I didn't fully verify it.

Image

Image

Image

Using the same process, I also made a reading-notes webpage. I happened to read Kazik's article today, so I used Designing Your Life as the topic — it's very beautiful and I really like it.

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

H

Image

H

Hy3也支持图片生成:

至于效果,ennnnnm,你们自评。

Image

Of course, you can also install Skills and try them out. Today, because I used it too frequently, I was already reminded about my quota limit.

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry