推文
760 tweets · 188 sources账号:全部mattpocockukdair_aisimonwbchernyadocompletetrq212dexhorthyyetonefelixrieseberg0xblacklightdani_avila7zarazhangruiAlchainHustbadlogicgamesdoteypetergyangvikingmuteryancarsonClaudeDevskunchenguidleon7haoctatedeviamzhihuikieranklaassenmckaywrigleylennysanthdxrRhysSullivangarrytanjakevin7johnlindquistkarpathyservasyy_aitheo0xMovez9hillsRLanceMartinaidenybaidingyihylarucodermitchellhmitsuhikomvanhornomarsar0steipetezeegGeminiAppGergelyOroszZaynHaoaddyosmaniantirezdanshipperdillon_mulroyelonmuskewind_devidoubiccnummanalirealWeZZardswyxthsottiauxyiliush0xPaulius0x_rodyAYi_AInotesAnthropicAIBarret_ChinaBenJames_____CMGS1988ChadMoranDanielMiesslerDaveJDavidKPianoDimillianFactoryAIFardeemMFarzaTVGoogleDeepMindHamelHusainHiTw93HilaShmuelIfanJewIndieDevHaileyJack_W_LindseyJackywineJasonZXJiaxi_CuiJinjingLiangKhalidWarsaLinearUncleLinghuaJMatthewBermanMaxForAIMeari_V2_0_GMnilaxQingQ77RaillyHugoSaccc_cTaniyatweets_VincentLogicai_explorer25aibuilderclub_alexalbert__alliekmillerandrewfarahantoinecojpanxue201arkuy99artmanasynkimobbssppllvvbearliubentossellbridgemindaibuaaxhmcailynyongyongcgtwtscharmaine_kleechrisbarberchrisparkXclairevodelba_oliveirademishassabisdoodlesteindriaforallelithrarelvissunericzakariassonfrancoisfleuretgabriell_labiamsahaj_xyzimdigitalashishimvihvjarredsumnerjerryjliu0jianshuojiayuan_jyjlongsterjonas_nellejoshalbrechtjpschroederkaiofreitaskepanolexruslogancyangmattypnabeelnifinetninthbit_aioran_gepatrickcpejmanjohnpetradonkaprathamgrvprathyvshquanruzhuoxiuquant_sheepraroquerealCaigurepsiacerohanpaul_aisamasamuelstroscheisawyerhoodshadcnshao__mengshaogefenhaoshreyansjtechgirl1908techwith_ramthorstenballtianyitobiaswuptoddsaunderstricaltturingoututuretomuniswap12untraceable_thevanillaCitronvelvet_sharkwengtianxinwquguruxiaogaifunxicilionycombinatorzachlloydtweetsziwenxu_zodchiii
@ctatedev·6月14日For You
I broke my "1-2 projects at a time" rule with Fable
Got a lot of exposure in a short amount of time and somehow burned through 15k requests 😅
What surprised me wasn't the benchmarks. A lot of benchmarks feel like noise tbh
It was watching it take a greenfield idea to ~90% complete on its own
And the output was actually good
The weirdest part was how each iteration felt. I'd throw an idea at it, wait a bit and then get to see what came back. Every time felt like opening a present
I'd have other models review the code and most of the time they found nothing
You could absolutely get to the same place with really good loops or by writing everything yourself
But it'd take way longer
What felt different was how I wasn't getting prototypes. I was getting products
It genuinely felt like the future
我打破了自己「同时最多 1-2 个项目」的原则，用了 Fable。 短时间内曝光量很大，不知不觉烧掉了 15k 次请求 😅 让我意外的不是 benchmark——说实话很多 benchmark 感觉都是噪音。 真正触动我的，是亲眼看着它把一个从零开始的想法独立推进到大约 90% 完成。 而且输出质量真的不错。 最奇妙的是每次迭代的体验——我扔给它一个想法，等一会儿，然后看到结果揭晓。每一次都像在拆礼物。 我还让其他模型来 review 代码，大多数时候它们什么问题都没找出来。 你当然也可以靠非常完善的 loop 或者自己全手写达到同样的地方。 但要花更长的时间。 感觉不同的地方在于：我拿到的不是原型，而是产品。 这真的让我觉得，未来已经到了。
@ctatedev·6月2日For You
Great advice
Bonus: global agent-browser starts ~150x faster than npx bc it runs the Rust binary directly
@andrelandgraf
Vibe coding tip: 1) Install agent-browser globally on your machine 2) Add this to your AGENTS.md dev workflow: After every change, verify all changes with agent-browser and fix any UI/UX issues. Do not stop until all changes have been verified.
很好的建议。 补充一点：全局安装 agent-browser 的启动速度比 npx 快约 150 倍，因为它直接运行 Rust 二进制，不走 npx。 【引用 @andrelandgraf】：Vibe coding 小技巧： 1）在你的机器上全局安装 agent-browser 2）在 AGENTS.md 的开发工作流中加上这一条： 每次改动后，用 agent-browser 验证所有变更并修复任何 UI/UX 问题，直到全部验证通过为止。
@ctatedev·5月15日For You
Introducing Zero
The programming language for agents.
I wanted a systems language that was faster, smaller, and easier for agents to use and repair.
Explicit capabilities. JSON diagnostics. Typed safe fixes.
Made for agents on day zero. https://t.co/uTrDOmyBR1
介绍 Zero——专为 agent 设计的编程语言。 我想要一门更快、更小、更易于 agent 使用和修复的系统语言：显式能力声明、JSON 诊断、类型安全修复。从第一天起就为 agent 而生。
@ctatedev·5月9日For You
Introducing zero-native
Build native desktop + mobile apps with web UI and Zig
→ Tiny binaries, low memory usage
→ Selectable web engines (WKWebView, WebKitGTK, WebView2, Chromium/CEF)
→ Next.js, Vue, Svelte, Vite, React
→ macOS, Linux, Windows, iOS, Android https://t.co/gMBinlXWOU
介绍 zero-native 用 Web UI + Zig 构建原生桌面和移动应用 → 极小二进制体积，低内存占用 → 可选 Web 引擎（WKWebView、WebKitGTK、WebView2、Chromium/CEF） → 支持 Next.js、Vue、Svelte、Vite、React → macOS、Linux、Windows、iOS、Android
@ctatedev·4月8日For You
agent-browser + Lightpanda + batch + multi-session is a dangerous combo
Completed this benchmark task in 5.85s https://t.co/Mkt4MgxWlV
@kunchenguid
alright agent nerds, if you care about your tokens and usage limits, pay attention to the tools you give to your agents. i built a benchmark that compared various browser tools for agents, and here's an example of their massive difference in cost and latency doing the same task https://t.co/pvuIo5TP44
agent-browser + Lightpanda + batch + 多 session 是个危险的组合（褒义）——用这套方案在 5.85s 内完成了 benchmark 任务 【引用】好了 agent 极客们，如果你在意 token 消耗和用量限制，注意你给 agent 配的工具。我做了一个 benchmark，对比了各种 browser 工具完成同一任务时在成本和延迟上的巨大差异。
@ctatedev·3月27日For You
Introducing the agent-browser dashboard
See exactly what your agent sees
→ Watch headless browser in real time
→ Manage all your sessions in one place
→ Debug with activity, console, network, and storage panels
agent-browser dashboard start https://t.co/4OyzTrrGA0
介绍 agent-browser dashboard 精确查看你的 agent 所看到的内容 → 实时监看无头浏览器 → 在一处管理所有 session → 通过 activity、console、network、storage 面板进行调试 agent-browser dashboard start [链接]
760 tweets · 188 sources