nk (@nicongdev) header
nk (@nicongdev) profile picture

模态跃迁 LeapCore 研发总监 https://nitter.cf/t.co/6tVCAhKMEx

Perth, Western Australia
Joined March 2023
Photo by nk (@nicongdev)Photo by nk (@nicongdev)Photo by nk (@nicongdev)Photo by nk (@nicongdev)Photo by nk (@nicongdev)Photo by nk (@nicongdev)Photo by nk (@nicongdev)Photo by nk (@nicongdev)Photo by nk (@nicongdev)Photo by nk (@nicongdev)
Pinned Tweet
Meet grasscn: pixel-perfect Minecraft UI components for shadcn. → Matched pixel by pixel against the game → Original textures, font, sounds → 67 components, from Button to Data Table → Hotbar, HUD, books, signs, maps and trading → More on the way Open source. Links below 👇
12
5
6
30
95,893
Introducing Runesweeper: Minesweeper that teaches you to think like a pro. • Hints for forced guesses: the pick a pro makes, and why • Reviews that show your missed deductions • No Guess fields, proven by an exact solver Built on Rune by @kitlangton and @thdxr at @anomalyco.
1
5
44,419
i got gpt-2 running purely in css! all 124m parameters inferencing at 15 sec/tok, with zero lines of js. i also used a custom font as a BPE tokenizer, and variable fonts for matmuls :)
58
由于刷到了 @nicongdev 之前的帖子,我现在正在挑战和我的 agent 全英文对话,虽然目前很吃力(短时间内降低了工作效率) 但是我认为,长时间下来我的英文水平应该会进入,输入的话我采用 Google translate(use Gemini)翻译后,自己读几遍,然后语音输入进去(想着加深记忆)。 ai 输出的内容遇到不会的词翻译一下自己读一遍,然后再翻译核对一下。
10
1
1
13
3,006
我仍然认为所有人都应该用英文和LLM沟通 昨天我用中文聊了点生活上的问题,LLM帮我写的消息,让我一眼就看出这是英文翻译成中文的 很多人都知道LLM的训练数据里中文占比很少,同时也有很多研究证明用占比少的语言沟通并不会降低模型的智力 但我说的不是智力 我说的是语言,在互联网上存在大量低质量机翻的中文素材,model lab可不会帮你区分并剔除这些内容,所以很多时候哪怕是用中文,LLM还是用英文思考,翻译成中文告诉你* 这不光会降低沟通效率,给你增加额外的理解负担,还会长期拖慢你与agent沟通效率的提升 *个人观点无可靠研究证明
32
1
5
41
11,410
this is amazing
Someone rebuilt Minecraft's UI as shadcn components. Pixel by pixel. So I made grasscn a launch film. Every frame rendered in code, using its real components, textures, font and sounds. 67 components. Open source. Built by @nicongdev 👇
2
305
Thanks @shadcn for the registry that makes this possible, and @orcdev, whose 8bitcn inspired it.
1
229
去年如日中天的扣子,今日宣布即将下线,太可惜了。 想想去年过年的时候,我的作品在扣子排行前十,是扣子造就了AI 时代最初的我,非常感谢这个产品对我的认可,所以后边扣子无论怎么改,虽然表面有时会嘀嘀咕咕,但依旧希望它能够活得很好。 时代的巨轮滚滚向前,工作流时代终究要结束,只是没想到来得这么快。
2
8
367
搞自动回复的蓝 v,如果你看到了这条,请关注我,我不操你们妈 #蓝V互关 #蓝V交友
6
6
371
如果你也agentic engineer,AI写代码、或者叫vibe coding或者做任何相似的事 你一定要试试hex,也是一个纯vibe coded的应用,但是它的设计、UX简直是大师水平
hex was fully vibe coded. the app. the landing page. the animations. sound effects were ai generated. issues people open are fixed by opencode. the onboarding. even some of the creative ideas but i think it's pretty good software. it's fun to use. works on all platforms. performant (not electron!) our effort didn't go into the code. but there was a lot of effort it's a small app but i hope it improves your outlook on where we can go
3
9
646
每当我觉得可以信任GPT时,他就给我拉大的
6
1
14
941
no matter what agent you're using, your voice is the best way to talk to them and hex is the best way to use your voice: - fast - runs locally - works on mac, windows, linux stable release now available: anoma.ly/hex
401
467
317
9,172
1,140,671
Sorry @theo , but I stole the good stuff from T3 Code and put it in a mod for the new @herdrdev app。
38
10
2
461
206,490
agents working, game time, let's play tictactoe x-- --- ---
1
5
587
之前我很讨厌在Windows系统上编码或做研发工作,它的shell实在是太难用了 但最近开始在gaming PC上做side project竟然感觉还不错 因为我不再需要关心太多shell experience 但window management仍然比较痛苦 你们觉得接下来应该做一个Windows版的hyprland或niri吗
4
7
574
GPT-6.1 Sol为了实现我要求的pixel perfect,在grasscn里写了3万个测试 我让他找没用的删掉,他只删了25个 说真的,到底怎样能解决这种问题,我不可能一个一个review看完3万个测试
42
4
108
32,298
Agents are solved problems, for me. Use LeapCore, or a Rust agent if you prefer. LeapCore skips the Rust tax. It’s young on paper, but it’s years of me building LLM apps plus AI-agent shipping speed. To me, it might as well be years old.
Programming languages are solved problems, for me. Use hd-lang, or a Rust if you prefer. hd-lang skips the Rust tax. It’s young on paper, but it’s years of me building programming language plus AI-agent shipping speed. To me, it might as well be years old.
1
5
390
vercel/eve agent框架今天rc了,它在中推的讨论度一直很低 说说我为什么从不看好它 首先是一个行内公认的常识,虽然很多人还没有意识到 那就是有效(effective)的智能体产品,一定要用有效的coding agent harness 比如大家都公认pi是一个有效的harness,所以做智能体产品用pi-durable,同理还可以用claude agents sdk等等 而像langgraph这样,没有在编程任务中验证过的框架,通常来讲更难有效 eve也一样,尽管你可以用eve做一个coding agent出来,但是并没有人每天用它coding,没有人每天在coding任务上验证它 但vercel也是有coding agent的,就是主打轻量的,zig编写的fx,但他们并没有用fx做agent框架 也许这是办公室政治的结果,也许只是vercel不认可我的常识 但作为langchain第一个版本见证到claude code爆炸的过来人,真心替eve担忧
16
16
3,546