Pinned Tweet
Which agent harness is worth learning from?
Not the one you can only use — the one you can read, run locally, and fully own.
Compaction, sandboxing, tool search, subagents — the designs the industry is converging on, in one open-source Go binary.
Search Agenvoy.
every office has one guy — big title, zero clue, bossing the experts around. everyone sees him as a clown.
all he does is show off and call pointless meetings. total waste of time.
then there’s the guy who’s always slammed, but look closer and it’s all busywork.
no deep expertise, zero efficiency — just a pile of little tasks that never add up to anything.
makes you wonder if he’s ever heard of the big rocks story.
if u think ai will replace engineers?
which one are you?
- selling an ai tool or a course
- staff eng who hasn’t shipped code in years, just leaves pr comments
- never built software in your life (or a junior who just started)
senior’s value isn’t writing less. it’s knowing what to skip.
$20 plan ships more than a junior on $200+.
what do you think?
Agenvoy is an open-source, self-hosted 24/7 personal AI agent that runs on your own computer — your memory, schedules, tools and credentials stay local.
Fills capability gaps itself — Creates, tests, and keeps a new tool when no suitable one exists, ready to reuse next time.
Shares one tool library across agents — Agenvoy, Claude Code, Codex, and other agents use the same sandboxed tools instead of rebuilding them.
A Taiwan-developed AI Agent Harness
GitHub: github.com/pardnchiu/Agenvoy
Make AI do the work on your computer—not just talk about it
從 2026 年初到現在,我的結論始終只有一句
算力租賃和 token 中轉,現在都賺不到錢,要賺錢去賣 API 給 Agent
算力租賃:拚不過專業資料中心
信用: 企業要的是 SLA 和合約保障,小平台連算力穩定度都很難讓人放心,更別說是否存在隨時倒閉的風險
設備: 小平台經常使用二手或上一代顯卡,每小時算力的性價比未必能與新世代資料中心設備競爭,硬體故障、功耗與維運效率也是個問題
隱私: 尤其是轉租第三方或個人閒置 GPU 的平台,客戶的資料和模型究竟運行在哪台機器上、由誰管理,平台難提供可靠的資料安全保障
這三個問題在第三方閒置 GPU 轉租平台上尤其明顯
企業客戶很難接受這樣的服務風險;一般開發者則有成熟的 GPU 雲端服務、OpenRouter、NVIDIA NIM、Ollama 等不同選擇,可以依照成本、便利性及部署需求解決問題
既沒有大型資料中心的穩定性,也沒有成熟平台的便利性,更沒有足以抵銷風險的價格優勢,使用者究竟為什麼要選擇這類小平台?
Token 中轉:拚不過專業玩家
單價: 拚不過中國低價中轉站,部分業者利用特殊額度、補貼或非正規管道取得低成本 Token,靠套利壓低售價,一般小平台很難與之競爭,更別說透過正規管道拿到的成本價,通常就要官方價的 75% 左右,扣掉金流手續費和維運成本後根本沒有降價空間
整合: 模型數量、API 相容性、路由能力和服務穩定性都拚不過 OpenRouter 等成熟平台,更別說品牌知名度與既有用戶基礎,甚至這件事本質只是單純包裝一層 API,幾乎沒有技術護城河
規模: 隨著官方 API 價格下降,Token 的轉售利潤持續受到擠壓。薄利就得靠量,但新平台缺乏用戶基礎,想擴大規模就得投入行銷與獲客成本,進一步消減本就有限的利潤
低價拚不過套利玩家,服務拚不過成熟平台,規模又不足以支撐薄利多銷。沒有獨特的供應鏈或技術優勢,單純做 Token 中轉,很難建立可持續的獲利模式。
利潤只能來自套利,而套利空間如今早已所剩無幾
但總有人看不清這點,以為這門生意好做
邱敬幃 Pardn Chiu retweeted
Agenvoy × NanoMD is here
NanoMD, the fastest JS Markdown parser I've built, now ships inside Agenvoy v1.2.2.
Richer, safer agent responses: Mermaid, KaTeX, Chart.js, all rendered on the fly.
邱敬幃 Pardn Chiu retweeted
One of the best Markdown-to-HTML renderers you'll ever see — renders a 4 MB file in just 88 ms.
- using in browser: load it from jsDelivr
- using in nodejs: install @pardnchiu/nanomd
邱敬幃 Pardn Chiu retweeted
Show code, not just words — my code is my pitch.
Taiwan · Infrastructure Engineering
邱敬幃 Pardn Chiu retweeted
总有人问我:我这个创业点子行不行?
说实话,点子是最不值钱的东西。
执行力强的人,99% 的点子都能做成小生意;
执行力差的人,给他台印钞机都能亏。
一人公司拼的从来不是想出多妙的点子,
是能不能把一件小事,坚持做到第 100 天。
Everyone so obsessed shipping fast these day
You, indie dev, what problem you solve today?
Any architecture thing finally click in your head?
I’ve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined.
I have really loved working at Ant and believe deeply in the mission, and most of all, the idea of a race to the top.
The idea that if we work hard and responsibly, we can raise the bar and inspire others to great outcomes. And if we all push each other to be better, things can be radically good. I will carry the spirit of this with me forever.
My colleagues at Anthropic are the most talented, thoughtful, competent, and earnest people I have ever met. I am glad they are at the wheel and trust them deeply.
To the community on X, thank you for everything. You’ve made this the most fun (and sometimes stressful) job I’ve ever had.
For me, it is time for a bit of a break before diving into something new.
If you want more excellent Claude Code content, you should follow @addyosmani , @lydiahallie, @_catwu, @ClaudeDevs and of course @bcherny who started it all.


























