Adam Cheung (@zhyd007) header
Adam Cheung (@zhyd007) profile picture

Software Engineer

Shanghai, China
Joined August 2017
Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)Photo by Adam Cheung (@zhyd007)
Nothing great was ever achieved without enthusiasm!
1
为了给之前做的交互简历配一套能随阅读状态动态切换的 BGM,我和 Claude 手搓了一套纯文本编曲语言,今天把它单独抽出来开源了:plainscore。 它和 MIDI 最大的区别就是:MIDI 是给机器读的二进制时间戳,而 .score 是给人读、给 AI 改的草稿纸。 在 plainscore 里你不用像写五线谱那样把每个和弦分解的音符抄八百遍: - 和弦(chords)和伴奏律动(part)是分开的,伴奏只写 `1 bass w`、`2& t2 q. soft` - 编排(form)就像调音台,直接写哪一小节谁进、谁退、音量多少 - 甚至支持 `ad lib.` 让竖琴按调式和概率随机点缀,自动避开主旋律 最爽的是在文件头塞了 20 行 Legend 注释,丢给任何 AI 都能秒懂语法,说一句「把第三段改成大提琴独奏」它改两行文本就完事,还能直接 git diff。 零依赖编译成 ABC 五线谱 + MIDI + WebAudio 事件流。 github.com/Innei/plainscore
1
44
现在 AI 模型训练 infra 除了 @daytonaio ,今天又多了解了一个 Harbor Framework harborframework.com/
40
这是 ChatGPT 的 bug 🐛吗 搜索功能无效
36
Adam Cheung retweeted
Your ChatGPT subscription is now also a Devin subscription
You can use any personal ChatGPT plan with Devin Just connect your Go, Plus, or Pro plan to Devin and all GPT model usage will draw down from you plan's quota
Readers added context they thought people might want to know
A ChatGPT plan can cover GPT model usage inside Devin for users with a separate Devin plan, but does not grant Devin access or act as a Devin subscription. docs.devin.ai/admin/billing/… help.openai.com/en/articles/20… devin.ai/blog/sign-in-w…
925
247
193
4,641
1,482,020
英语学习小助手
GPT-6 and Intelligent UI, now rolling out in ChatGPT for everyone. Intelligent UI in ChatGPT delivers fast, interactive answers that make everyday questions more visual, complex topics easier to grasp, and interactive tools for your task available on the spot.
1
30
Pro 用户有 $100 usage credit 活动,一旦兑换成功,也会先消耗 credit
刚刚领了 $200 的 Claude 的 API Credit;本以为是 Anthorpic 大发善心。实际上是他把 Agent SDK 的支付拆出去了;后续不能走订阅了,只能走 API;不能薅羊毛了。。。 来源见评论区
34
Railway 试用到期了,换到 @vercel 平台部署 Golang 服务 oss-criticality-radar.vercel…
Claude Code 对开源项目重要性评分达到 .4 的开发者提供 6 个月的 Claude Max 20x 对于重要性评分使用 Golang 写的项目本地测试:github.com/ossf/criticality_… 这两天用 @grok 最新的 4.5 模型搭了一个 Web 界面:oss-criticality-radar.vercel… 核心算法还是那个 Golang 服务,不过托管到 @Railway 平台了
31
Smart
something fun i made this weekend
63
Adam Cheung retweeted
something fun i made this weekend
522
265
241
8,416
5,869,387
有很多功能 Tiptap 已经做好了
15 hours + $500 cost. From idea to production. 昨晚到现在,Agent 连续工作了 15 小时,做出一个类似 Claude Docs 的产品,现在已经上线。 我没看代码,也没读文档。只看 summary、设定目标、做架构和产品决策。 Rust + Tiptap + Yjs,完全托管在 Cloudflare。Notion 式编辑、AI 与多人实时协作、OAuth、MCP、Google 登录,都跑通了。 AI 时代,写代码的成本正在迅速下降。真正拉开差距的,是架构设计、产品取舍和 Taste。
55
给自己的 pet project 做性能优化(JavaScript)时才会注意到的几点: 1. Map/Set 用数字类型做 key 时比用字符串或对象操作起来更快 2. Array.from(frozen) 比用 slice() on a frozen array 更快 3. typed array 这种数据结构在性能优化算法里挺常用
1
2
95
这两个案例体现了理论算法复杂度中的常数因子在实际应用中的重要性
We asked ten Claude Opus 5.5 agents to devise a faster shortest-path algorithm and prove it in Lean. Within 15 hours, they produced C-HD: a formally verified improvement over the published bounds.
Readers added context they thought people might want to know
C-HD's formally verified bound improves priors only in a narrow sparse density range (m ≤ n⌊⌊log₂ n⌋^{3/4}⌋) as an asymptotic guarantee with enormous constants and no practical speedup demonstrated. vals.ai/blogs/faster-s… github.com/spicylemonade/…
57
ChatGPT Images 2.5 帮我设计了一个协同编辑的工作空间 这个 Happening Now 板块不错,工程实现上实现难度应该不小
2
122
Fable 5.1 Max 可以生成高质量的英语听说学习资料 试了一下只有高级人工智能模型才能做到,当然使用成本也高 听力资源来自 @BBC
1
130
Adam Cheung retweeted
what if copy/paste was smart? powered by @typesafeai jev it feels like every computer interaction will get rewritten
327
564
248
9,806
1,097,134
前司有个业务需要对邮件根据来信进行分类,要么 A 要么 B 感觉 Jev 挺适合这个业务
55
Adam Cheung retweeted
New experiment: json-render + jev The future Generative UI is instant Your components, your actions, your design system Rendered in milliseconds
223
507
265
7,789
1,349,160
Adam Cheung retweeted
infinite faq
291
266
227
6,505
1,367,495