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
@kunchenguid·6月18日For You
time to reveal the next big gun in my agentic engineering setup - Firstmate!
it's the only agent session i directly talk to now. if you find it mentally exhausting to juggle between all the agent sessions, this is the solution
free & open source as always - details below 👇 https://t.co/UEvPoPv3iK
揭晓我 agentic engineering 工作台的下一件重器——Firstmate！ 这是我现在唯一直接对话的 agent session。如果你觉得同时管理多个 agent session 很耗费心力，这就是解法。 一如既往免费开源——详情见下 👇
@kunchenguid·6月9日For You
oh gosh, this is kind of a big deal
DO NOT ASK YOUR AGENTS TO DO TDD!
i now have empirical evidence that Test Driven Development is harmful for coding agents
what other popular skills do you want me to debunk?
details about the evaluation below 👇 https://t.co/UsRSek5mB0
哦天哪，这事挺重要的—— **不要让你的 agent 做 TDD！** 我现在有实证数据表明，测试驱动开发对 coding agent 是有害的。 还有哪些流行的实践你想让我来打假？ 详细评测结果见下方 👇
@kunchenguid·5月22日For You
i'm strongly against model companies focusing too much on harness, but i would love to hear if anyone has a strong argument for it
my reason against it:
if openai didn't build GPT 5.5, no one else can. this is their core competence
if openai didn't build codex cli and app, we have opencode and t3code. building harness is NOT their core competence
this is not saying products like claude code, codex aren't good - i genuinely think these are top tier products built by really talented people
my point is - the world might be a better place if model companies focus more on their core capability and give us better, faster, safer and cheaper models, rather than competing with the ecosystem in the application layer
what do you think?
@gdb
the model alone is no longer the product
我强烈反对模型公司把太多精力放在 harness 上，但如果有人有充分的反驳理由，我很想听。 我反对的理由： 如果 OpenAI 不做 GPT 5.5，没人能做——这是他们的核心竞争力。 如果 OpenAI 不做 Codex CLI 和 App，我们有 opencode 和 t3code。做 harness **不是**他们的核心竞争力。 这不是说 Claude Code、Codex 这些产品不好——我真心觉得这些是由极其优秀的人做出来的顶级产品。 我的观点是：如果模型公司更专注于核心能力，给我们更好、更快、更安全、更便宜的模型，而不是在应用层与生态竞争，这个世界可能会更好。 你怎么看？ > 引用 @gdb：模型本身已经不再是产品。
@kunchenguid·5月17日For You
i never use “remote control” in agent harnesses because they feel more like middle grounds
i tailscale to my mac + use terminus to do the real thing. this setup has no extra cost https://t.co/I9sJT1Ot0W
我从不在 agent harness 里用「remote control」，因为感觉更像是个不上不下的折中方案。 我直接 Tailscale 连到 Mac，然后用 Terminus 做真正的操作。这套方案零额外成本。
@kunchenguid·5月12日For You
time to reveal my HTML workflow with agents
HTML is the new markdown. Lavish is the new editor for your HTML artifacts
just tell your agent - discuss the technical plan with me using `npx lavish-axi`
100% open source and runs locally. details in thread below 👇 https://t.co/TUEhT3kN0L
揭秘我的 agents HTML 工作流 HTML 是新的 markdown。Lavish 是你 HTML artifacts 的新编辑器。 只需告诉你的 agent——用 `npx lavish-axi` 和我讨论技术方案 100% 开源，本地运行。详情见下方 thread 👇
@kunchenguid·5月10日For You
sharing some real data from me using agents on real projects. all my code changes are done with either opus 4.7 or gpt 5.5
a whopping 68% of the changes made by agents had a mistake in it, and got saved by no-mistakes
biggest source of gaps was a change made without updating related documentation, followed by problems caught via code review
i now can't imagine how i could have kept my codebases in order if i didn't have no-mistakes - they would absolutely have turned into a sloppy mess even with the best models we have today
分享一些我在真实项目中使用 agent 的真实数据。所有代码变更都由 opus 4.7 或 gpt 5.5 完成。 高达 68% 的 agent 变更存在错误，全靠 no-mistakes 工具救场。 最大的问题来源是：改了代码但没更新相关文档，其次是 code review 阶段发现的问题。 现在我无法想象，如果没有 no-mistakes，我的代码库会变成什么样——即使用今天最好的模型，代码库也会绝对沦为一团糟。
@kunchenguid·5月8日For You
the new @OpenAI realtime voice model just released + gpt 5.5 fast mode brings us a new possibility -
realtime speech to live presentation!
i just talk, and the whiteboard would whiteboard itself
prototype is open sourced. details in thread below - https://t.co/vqVTdvqoQL
OpenAI 新发布的实时语音模型 + gpt-4.5 fast 模式带来了一个新可能—— 实时语音转现场演示！ 我只需要说话，白板就会自动生成内容 原型已开源，详见 thread
760 tweets · 188 sources