Ben Davis of Illuminate is an SF treasure
来自 Illuminate 的 Ben Davis 是 San Francisco 的瑰宝
GT
Garry Tan
@garrytan ↗President & CEO @ycombinator —Founder @garryslist—Creator of GStack & GBrain—designer/engineer who helps founders—SF Dem accelerating the boom loop
3 最新130 累计48 期
每条推文 hover 显示单独 ▶
If you haven’t seen 7x7 at SF City Hall it is a sight to behold. Go see it before it goes away July 4th
如果你还没在 SF City Hall 看过 7x7，那真是一处值得一看的景象。趁它在 7 月 4 日撤展之前去看看吧
The California asset seizure tax is only going to impoverish the state and reduce funding for health care The SEIU UHW union doesn’t care if quality of care or even the state’s ability to care for its people goes away There is only pure self interest there
California 的资产没收税只会让这个州变得更贫困，并减少对医疗保健的资金支持。SEIU UHW union 根本不在乎护理质量，甚至也不在乎这个州照顾其人民的能力是否会消失。那里有的只有纯粹的自身利益
This is why I made GBrain and open sourced it
这就是我为什么做了 GBrain 并将它 open source（开源）。
I think one underestimated thing when we look back on it was how useful it is to have your own personal brain and company brain in 2026 at the dawn of usable AGI AGI gives you the intelligence You still have to collect your personal context to get the real unlock
我认为，当我们回过头来看时，一个被低估的点是：在 2026 年、可用 AGI 黎明到来之际，拥有你自己的 personal brain（个人大脑）和 company brain（公司大脑）是多么有用。AGI 给你 intelligence（智能），但你仍然必须收集属于你个人的 context（上下文），才能真正实现那次 unlock（解锁）。
Tip: pick the worst thing you are afraid to show your board for your next board meeting and make it slide one (you can only do this with good boards btw) Make it a big deal you did that. Make it a habit.
提示：在下一次 board meeting 上，挑一件你最害怕拿给董事会看的、最糟糕的事情，把它做成第一张幻灯片（顺便说一句，这只有在你有一个好的 board 时才做得到）。要把你这么做这件事当成一件大事来强调。并把它变成一种习惯。
AP reports on the $200M sales level Asset seizure is evil. My grandparents and parents generations came to America to flee asset seizure and this ideology. We came here to build prosperity. An America that still builds is worth defending.
AP 报道了以 2 亿美元销售额为门槛的 Asset seizure（资产没收）问题。Asset seizure 是邪恶的。我祖父母和父母那几代人来到 America，就是为了逃离 asset seizure 和这种意识形态。我们来到这里，是为了创造繁荣。一个仍然在建设中的 America，值得被捍卫。
Bernie Sanders introduced a bill to seize 50% of any AI startup that crosses $200M in revenue. The same anti-prosperity bloc spent the year trying to ban startup acquisitions, blocking the only exit 85% of founders ever get. This is a war on building startups in America.
Bernie Sanders 提出了一项法案，要没收任何营收超过 2 亿美元的 AI startup（AI 初创公司）50% 的股权。同一个反繁荣阵营这一年里一直试图禁止 startup acquisitions（初创公司并购），堵死了 85% 创始人唯一能获得的退出路径。这是在向 America 的创业建设开战。
For people who don’t have a clear sense of the future they want, AI is just another mechanism of control But in the hands of someone with agency, AI is the breaker of chains, something that lets you do things no humans can do alone. AI can be a liberator if you choose agency.
对于那些并不清楚自己想要什么样未来的人来说，AI 只是另一种控制机制；但在有 agency（主导力、能动性）的人手中，AI 是打破枷锁的工具，是一种让你做到单靠人类无法完成之事的东西。如果你选择 agency，AI 就可以成为解放者。
Rough estimate on $ productivity lost by Fable 5 ban: $12M per hour Frontier AI-coding daily actives, mid-2026: 5M devs Fully-loaded cost: $90/hr Work routed to Fable in 48 hours: 17.8% Fable is on average ~15% more productive Effective throughput loss per dev = 17.8% × 15% ≈ 2.7% of output 2.7% × $90/hr = ~$2.40/dev/hr × 5M devs = $12M per working hour
对 Fable 5 封禁所造成生产力损失金额的粗略估算：每小时 $12M。到 2026 年中，Frontier AI-coding 的日活跃用户：500 万开发者；fully-loaded cost（完全成本）：$90/小时；在 48 小时内被路由到 Fable 的工作：17.8%；Fable 平均可提升约 15% 的生产力；因此每位开发者的有效吞吐损失 = 17.8% × 15% ≈ 产出的 2.7%；2.7% × $90/小时 = 约 $2.40/开发者/小时；再乘以 500 万开发者 = 每个工作小时 $12M。
I think you still need both, but the main lede is: technical founders now have access to business thinking Business founders now have access to technical thinking Net net: more startups that actually work, period
我认为你仍然两者都需要，但主要的 lede（核心信息）是：技术型创始人现在能够获得商业思维；商业型创始人现在能够获得技术思维；net net（归根结底）：真正能跑通的 startup 会更多，就这么简单。
We don’t care what your age is We care if you can build with craft and care And if you can make something people want
我们不在乎你的年龄是多少；我们在乎的是你是否能以精湛技艺和认真态度去 build（构建）；以及你是否能做出人们真正想要的东西。
9Mothers is one of those technologies whose impact is immediately obvious to warfighters
9Mothers 是那种其影响对 warfighters 来说一眼就很明显的技术之一
You’ll never achieve anything if you are afraid of being cringe Then again this is from someone trying to use looksmaxxing to achieve outsize return, good luck with that
如果你害怕显得 cringe，那你将永远一事无成。话又说回来，这话出自一个试图靠 looksmaxxing 获得超额回报的人，祝你好运吧
Which leads to… Bowen’s differentiation of self Low differentiation = highly reactive to others' emotional cues, loses self in relationships, can't hold boundaries under pressure — collapses into "we-ness" (everyone must feel the same or the bond feels threatened). Two flavors of the same disease: chameleons (adjust themselves to please, fusion mode) and bullies (force others to conform, the asshole mode). Both are undifferentiated. The self-abandoner and the asshole are the same person. The fix is the same: lower the chronic anxiety, raise the capacity to stay a separate self in contact.
这就引出了…… Bowen 的 differentiation of self（自我分化）。低分化 = 对他人的情绪线索高度反应，在关系中失去自我，在压力下无法守住边界——坍缩成“we-ness”（我们一体感：每个人都必须有相同感受，否则这段联结就会显得受到威胁）。同一种病有两种表现：变色龙型（为了取悦他人而调整自己，融合模式）和 bully 型（强迫他人顺从，asshole 模式）。两者都属于未分化。那个放弃自我的人和那个 asshole，其实是同一种人。解决办法也一样：降低长期性焦虑，提高在保持接触的同时仍能作为一个独立自我存在的能力。
Open source is the escape hatch for businesses to be able to continue to control their own destiny long term
Open source（开源）是企业得以长期继续掌控自身命运的逃生舱口。
The next generation of young people who change the world will almost certainly be the people who are most adept at making long-running multi-stage multi-team agent tasks work extremely well, and at high volume and across every part of their personal and work lives.
下一代改变世界的年轻人，几乎可以肯定会是那些最擅长让长期运行、跨多个阶段、涉及多个团队的 agent（智能体）任务以极高效果运转起来，并且能在大规模下将其应用到个人生活和工作生活各个方面的人。
In AI most people are still trying to use old maps on a new territory. Throw the maps away. It's time to draw new ones. The only way you can do it is walking the land.
在 AI 领域，大多数人仍在试图用旧地图去理解一片新大陆。把那些地图扔掉吧。现在是时候画新的了。而你唯一能做到这一点的方法，就是亲自走过这片土地。
I wish the significance of the model came from more people actually using it and coming to their own conclusions But yes most people in the world learn about it through signifiers and not through interacting with that which is signified
我希望这个 model 的意义，来自更多人真正去使用它，并得出他们自己的结论。不过，确实，世界上大多数人了解它，是通过 signifier（能指）来了解，而不是通过与被 signified（所指）的对象直接互动。
So proud of @datacurve (YC W24) - building THE defining software engineering benchmark in DeepSWE Tired? SWE-Bench Pro Wired? Datacurve DeepSWE
为 @datacurve（YC W24）感到无比自豪——他们正在打造 DeepSWE 中具有定义性的软件工程 benchmark（基准测试）：Tired? 用 SWE-Bench Pro。Wired? 用 Datacurve DeepSWE。
I just discovered forceBlockStreamingForReasoning = resolvedReasoningLevel === "on" for OpenClaw and frankly I love it Seeing the reasoning traces of my claw with Claude Fable 5 is a mind-blowing experience. Seeing the tale of the tape, and being able to engage with ideas at a deeper level has always been my catnip.
我刚发现，对 OpenClaw 来说，forceBlockStreamingForReasoning = resolvedReasoningLevel === "on"，坦白说我太喜欢了。用 Claude Fable 5 看到我的 claw 的 reasoning traces（推理轨迹），真是一次震撼心智的体验。看到完整过程，并且能够在更深层次上与想法互动，一直都是最让我着迷的东西。
Everyone thinks AI coding tools set founders free. Watch what people actually build with them: rules, approvals, process, layers. The same cage, assembled faster. The tool that can scaffold anything in an afternoon will scaffold your bureaucracy in an afternoon too. Speed of construction is speed of calcification. Build the thing that lets you create new things: experiences that didn’t happen before.
每个人都以为 AI coding tools（AI 编码工具）让创始人获得了自由。可看看人们实际上用它们构建了什么：规则、审批、流程、层层叠加。还是同一个笼子，只是组装得更快。那个能在一个下午 scaffold（快速搭建）任何东西的工具，也同样能在一个下午 scaffold 出你的官僚体系。建造速度，就是固化速度。去构建那种能让你创造新事物的东西：那些以前从未发生过的体验。
The centerpiece stat is a self-own: she cites a 35-year study where 12.3% of 677 gifted kids reached "eminence" (full professor, Fortune 500 exec, federal judge) as proof the category is a lie. The general-population base rate for outcomes that rarefied is far worse, much closer to zero than double digit percentage. 12.3% is the selection mechanism working spectacularly.
最核心的那个统计数据反而是在自打脸：她引用了一项为期 35 年的研究，称在 677 名 gifted kids（天才儿童）中有 12.3% 达到了“eminence”（杰出地位：full professor、Fortune 500 exec、federal judge），并把这当作“gifted”这个类别是个谎言的证据。可对普通人群来说，这种极其稀有结果的基准发生率要低得多，更接近于零，而不是两位数百分比。12.3% 恰恰说明这种筛选机制发挥了惊人的作用。
The community note clearly specifies the writer Katie Arnold-Ratliff is the one whose understanding of gifted is inherently flawed Stop making it cool for writers to in-group signal with lies
那条 community note（社区注释）已经明确说明，真正对 gifted（天才）概念存在根本性误解的人，是作者 Katie Arnold-Ratliff。别再把靠谎言进行圈内身份信号传递，包装成一种很酷的事了。
Gavin Newsom came for a @garryslist event at YC and we were proud to talk little tech, YIMBY, how to stop the the asset seizure tax, and how to keep California's innovation wave going while making sure things work for every Californian. Thank you @GavinNewsom
Gavin Newsom 来参加了 YC 的一场 @garryslist 活动，我们很高兴能聊聊一点 tech（科技）、YIMBY、如何阻止 the asset seizure tax，以及怎样让 California 的创新浪潮持续推进，同时确保这一切对每一位 Californian 都真正有用。谢谢你，@GavinNewsom。
May common sense reign in San Francisco for 100 years Aaron Peskin, enjoy being a private citizen.
愿常识在 San Francisco 长存 100 年。Aaron Peskin，祝你享受作为普通公民的生活。
Performative nonprofit industrial complex must be rooted out and defunded. Their political cronies and grifter friends must not be allowed to squander the gifts of San Francisco any longer.
必须铲除并停止资助这种作秀式的 nonprofit industrial complex（非营利工业复合体）。绝不能再允许他们的政治同伙和骗子朋友继续挥霍 San Francisco 赋予的一切。
Nessie just became the best way to get all your existing context, memory and history from ChatGPT, Perplexity, and Gemini into all the other places you have memory, and also get it into OpenClaw/Hermes Agent. Their OpenClaw and MCP servers are ace.
Nessie 刚刚成为把你在 ChatGPT、Perplexity 和 Gemini 中已有的所有 context（上下文）、memory（记忆）和 history（历史记录）导入到其他所有支持 memory 的地方的最佳方式，也能把这些内容导入 OpenClaw/Hermes Agent。他们的 OpenClaw 和 MCP servers（服务器）非常出色。
Fable 5 is the biggest model energy I've ever seen
Fable 5 是我见过“模型能量”最强的一次。
Just trying to work on fixing GStack and running into this with Fable 5 *long sigh*
我只是想修一下 GStack，结果又在 Fable 5 这里撞上这个问题了 *长叹一口气*
I eat Aaron Peskin's hate for breakfast. Guys, we're winning. It feels good. Aaron Peskin is no public servant. He was all along just a grifter trying to protect his views at the expense of the people. We are the real people powered movement of San Francisco.
我把 Aaron Peskin 的仇视当早餐吃。各位，我们在赢。这感觉真不错。Aaron Peskin 根本不是什么公仆。他从头到尾都只是个 grifter，一心想牺牲民众利益来维护他自己的立场。我们才是 San Francisco 真正由人民驱动的运动。
Flock Safety makes cities safer Stop protecting criminals
Flock Safety 让城市更安全，别再保护罪犯了
NIMBYism only impoverishes the people but people like Connie Chan will say or do anything to get political power
NIMBYism 只会让民众变得更贫困，但像 Connie Chan 这样的人为了获得政治权力，什么话都说得出来，什么事都做得出来
Because this is a brand new form of centrism being born in San Francisco The 2030’s will look back on this time when the new San Francisco common sense Democrat was born from the failures of the hard left
因为这是一种正在 San Francisco 诞生的全新中间主义形式；到了 2030 年代，人们回看这个时期时，会把它视为新的 San Francisco 常识派 Democrat 从强硬左派的失败中诞生的时刻
Educating people on how to use the AI tools has become a serious bottleneck
教人们如何使用这些 AI 工具，已经成为一个严重的瓶颈
GBrain v0.42.30 can now give you a detailed summary of how your thinking has changed over time.
GBrain v0.42.30 现在可以为你详细总结你的思考方式是如何随着时间推移而变化的。
We never said we don't upload any user data to the cloud. We said the code (file contents) specifically doesn't. Paxel is here to help you, and over time as local models get better, we'll be able to do even more locally. Can't wait, tbh.
我们从未说过我们不会把任何用户数据上传到云端。我们说的是，code（代码；文件内容）这一项明确不会。Paxel 的存在是为了帮助你，而且随着 local models（本地模型）不断变得更好，我们以后将能在本地完成更多事情。说真的，我已经等不及了。
I also want to help people become more legit with Paxel
我也想帮助人们借助 Paxel 变得更 legit（更正规 / 更靠谱）。
Local government matters Oakland mismanagement is fixable but there has not yet been a resurgence of common sense the way SF had it @empoweroak is working on this
地方政府很重要。Oakland 的管理失当是可以纠正的，但目前还没有出现像 SF 那样的常识回归，@empoweroak 正在为此努力。
GBrain gives OpenClaw and Hermes Agent wings
GBrain 为 OpenClaw 和 Hermes Agent 插上了翅膀
You can now finally try one of those big projects I was teasing that I started working on a few months ago. We want this to over time get better at helping you learn the best techniques to build better software, faster.
现在你终于可以试试我之前一直在卖关子、几个月前就开始着手的那个大型项目之一了。我们希望它随着时间推移，能更擅长帮助你学习那些能让你更快构建更好软件的最佳技术。
This is the definition of must-have for the future drone war.
这就是未来 drone war 的必备定义。
Two YC decacorns in one day and one of them is building commercial fusion. Polaris hit 150 million degrees C, first privately funded machine to do it. This is the abundance future, built by people who actually ship.
一天之内出现了两家 YC decacorn（估值超百亿美元公司），其中一家还在打造商业聚变。Polaris 达到了 1.5 亿摄氏度，是第一台做到这一点的 privately funded（私人资助）机器。这就是由那些真正能把产品做出来并交付的人所构建的 abundance（丰裕）未来。
So close to product market fit is still not product market fit
非常接近 product-market fit（产品市场契合）依然不等于真正达到 product-market fit（产品市场契合）。
[1]
The Pelley Firing Is Not a Free-Press Tragedy. It’s an Autopsy.
Pelley 被解雇并不是一场新闻自由的悲剧，而是一份验尸报告。
San Francisco is healing. We don't fall for this type of ridiculous gaslighting anymore.
San Francisco 正在痊愈。我们已经不会再被这种荒谬的 gaslighting（煤气灯操纵）所欺骗了。
Thank god Phil Kim beat the insane woman named Brandee who literally doxxed my home address and made me sell my home and move my family to a new one
谢天谢地，Phil Kim 击败了那个叫 Brandee 的疯女人；她居然把我家的住址开盒（doxxed）了，逼得我卖掉房子，把家人搬去新的住处。
tfw you land many PRs and have many more to go before you sleep
当你提交并合入了很多 PR（pull request），但在你休息之前，还有更多 PR 要继续处理。
GBrain is the agentic swiss army knife for retrieval and memory
GBrain 是一把面向 retrieval（检索）与 memory（记忆）的 agentic 瑞士军刀。
We have 16 partners funding 40 to 60 companies per year, usually amounting to more than 50 to 200 new SF residents per partner per year (depending on how many cofounders and how many already live here, but multiply by 16 and it’s a lot!)
我们有 16 位 partners（合伙人），每年资助 40 到 60 家公司，通常这意味着每位 partner 每年会带来超过 50 到 200 名新的 SF 居民（取决于有多少 cofounders（联合创始人），以及其中有多少人原本就住在这里，但乘以 16 之后，这个数字就非常可观了！）
This sounds like a small thing but it's big. Platforms need to stay open and it shouldn't require a lot of work to get your data. Because where the AI harness wars are going... someone else's harness is just going to be you sharecropping someone else's AI ecosystem.
这听起来像件小事，但其实很重要。平台需要保持开放，而且获取你的 data（数据）不应该需要花很多功夫。因为 AI harness wars 将走向何方……别人的 harness 终究只会意味着：你在别人的 AI ecosystem（生态系统）里当 sharecropper（佃农）。
You should want to control and host your own memory It’s the one thing that you should be able to take to any platform Watch for this to be a defining battle in the new browser war: the AI harness wars of 2027
你应该希望控制并托管你自己的 memory（记忆）。这是唯一一件你应该能够带到任何平台上的东西。注意，这将会成为新一轮 browser war（浏览器战争）中的决定性战役：2027 年的 AI harness wars。
Unfortunately the building boom has not hit San Francisco yet and that's not a good thing
不幸的是，建筑热潮尚未席卷 San Francisco，而这并不是一件好事
Don’t forget to vote by June 2. Use my guide to see how we can fix California garrysguide dot org
别忘了在 6 月 2 日前投票。用我的指南看看我们如何修复 California：garrysguide dot org
This is the Wolff we need on our side Vote Patrick Wolff for insurance commissioner
这才是我们需要站在我们这一边的 Wolff。投票支持 Patrick Wolff 竞选 insurance commissioner
A founder kept saying "if only we had money we'd do X." Money is not the fire. Money is gasoline you pour on a fire that already exists. You don't have a funding problem. You have a "people don't want it yet" problem. Go make the first fire.
一位 founder 一直在说：“要是我们有钱，就能做 X。” 钱不是火种。钱是浇在一团已经存在的火上的汽油。你不是有 funding（融资）问题。你的问题是“人们还不想要它”。先去把第一团火生起来。
Virtue signalers who don’t actually care about the outcomes for the students of the UC system really should have had no part in the UC Regents Unfortunately they have damaged beautiful institution and their names should be remembered for it
那些实际上并不关心 UC system 学生结果、只会做 virtue signaling（道德姿态表演）的人，本来就不该参与 UC Regents。不幸的是，他们已经伤害了这个美好的机构，而人们应该因此记住他们的名字。
Anti Asian racism alive and well in University of California admissions
反 Asian（亚裔）种族主义在 University of California 的 admissions（招生）中依然真实存在，而且状况良好。
Founders must stop trying to building 2010-era businesses with 2026-era technology. Don't try to rebuild Foursquare or Yelp. Don't try to recreate Basecamp by 37 Signals with $10/mo SaaS pricing. Don't underprice! If it works it's worth a lot more. Don't be tempted to become "Tech enabled PE" with revenue tricks. The rules of tech changed with AI. Play the new game.
创业者必须停止用 2026 时代的技术去做 2010 时代的生意。不要试图重建 Foursquare 或 Yelp。不要试图用每月 $10 的 SaaS 定价去复刻 37 Signals 的 Basecamp。不要把价格定得太低！如果它真有用，那它值的钱会多得多。也不要被诱惑去做那种靠营收花招运作的“Tech enabled PE”。随着 AI 的出现，科技行业的规则已经变了。去玩这场新游戏。
GStack is still getting better daily. Just shipped v1.47 Thanks @jayzalowitz for the PR idea Extra useful for when I need to fix GBrain issues and need to create a precise spec about the bug directly out of the OpenClaw instance that hit the issue
GStack 现在仍然每天都在变得更好。刚刚发布了 v1.47，感谢 @jayzalowitz 提出的 PR 想法。这个功能特别有用，尤其是在我需要修复 GBrain 问题，并且要直接从触发问题的 OpenClaw 实例中，为这个 bug 写出一份精确 spec（规格说明）的时候。
Saikat Chakrabarti said that a single member of Congress from San Francisco is "useless." This type of statement should immediately disqualify him from office. He doesn't want to represent the people. He just wants to buy personal power and clout. He is no public servant.
Saikat Chakrabarti 说，来自 San Francisco 的一位国会议员“毫无用处”。这种说法本身就应该立即让他失去担任公职的资格。他并不想代表人民。他只想买来个人权力和影响力。他根本不是什么公仆。
Idea fusion is fun, and idea fusion with LSD (lateral synaptic drift) is even more fun
创意融合很有趣，而把创意融合与 LSD（lateral synaptic drift，侧向突触漂移）结合起来就更有趣了
Exa is what I trust for all my agents. We use it at YC. We use it in all my OpenClaw and Hermes Agents. There is no other option that is as fast, as reliable, and as complete. When your agents need to search the web, accept no substitutes.
Exa 是我为所有 agent 都信赖的工具。我们在 YC 使用它。我们也在我所有的 OpenClaw 和 Hermes Agents 中使用它。没有其他选择能同时做到这么快、这么可靠、这么完整。当你的 agent 需要搜索 web（网络）时，不要接受任何替代品。
The benefits of AI are real now, but we will never get there in the United States (and America will cede its leadership) if we can't solve the bigger media issue of a mass smokescreen by forces that want to destroy America
AI 的好处现在已经是真实可见的，但如果我们无法解决一个更大的 media（媒体）问题——那些想要摧毁 America 的力量制造出的大规模烟幕——那么我们在 United States 永远无法真正实现这一点（而且 America 也会让出其领导地位）
Still absolutely amazed I got to be on a pod with Rick Rubin right after Clown from Slipknot Code can be art too
仍然完全不敢相信，我居然能在播客里紧接着 Slipknot 的 Clown 之后和 Rick Rubin 同台。代码也可以是艺术。
Weird quirk of history: I worked at Microsoft in 2003-2005 when they were trying to make WinFS work I guess with LLMs, WinFS can happen now, and it's called GBrain
历史里一个很奇怪的小插曲：我在 2003 到 2005 年在 Microsoft 工作，那时他们正试图让 WinFS 跑起来。我想，有了 LLM（大语言模型），WinFS 现在终于可以实现了，而它叫作 GBrain。
OK I guess we're going pretty fast improving GBrain day to day right now
好吧，我想我们现在确实在以很快的速度日复一日地改进 GBrain。
Also 28 bug fixes landing in a single bug fix wave that rolls up 22 community PRs and 14 issues
另外，一次单独的 bug 修复波次中就合入了 28 个 bug 修复，汇总了社区的 22 个 PR 和 14 个 issue。
Full eval report and fixtures published and open source here Any memory system is welcome to run theirs against these evals and I'll publish them alongside.
完整的 eval（评测）报告和 fixtures（测试夹具）已在这里发布并开源。欢迎任何 memory system（记忆系统）用这些 evals（评测集）来跑他们自己的结果，我也会把这些结果一并发布。
Ken Griffin doesn’t understand the ceiling just got raised. Some 20-something maybe reading this will build the cracked AI-human-computer-symbiosis team that will supersede his whole operation because he is too distracted about lowering cost Boil the ocean don’t cut your costs
Ken Griffin 还没有意识到，上限已经被抬高了。某个二十多岁、也许正读到这段话的人，会组建出那个虽不完美但已打通的 AI-human-computer-symbiosis 团队，并全面超越他的整套业务，因为他还过于分心于压低成本。要去 boil the ocean（放手做大事、解决宏大问题），不要一味削减成本
[1]
For personal AI scenarios against my 120k markdown brain ZeroEntropy has earned the top slot
在面向我这个 120k markdown brain 的个人 AI 场景中，ZeroEntropy 已经赢得了第一的位置
Saturday morning and it’s a good time to think a bit about how our functional systems are being torn down by a mind virus by two philosophers: Foucault and Derrida Foucault: His framework tells you that every institution claiming to know something is really just exercising power. Medicine, engineering, law, science. Apply that at civilizational scale and you get exactly what Dan Wang warns about: a society that lost the will to build. Process knowledge — the tacit know-how that only exists in the hands of people who actually make things — dies when a culture decides that all knowledge claims are suspect. America went from building the Interstate Highway System and the Apollo rockets to being unable to build a train from LA to SF. That didn't happen because we forgot the engineering. It happened because we built an entire intellectual class whose job is to interrogate every system rather than improve one. Derrida: His move is that every commitment contains its own contradiction, so you can never land on firm meaning. Run that as societal firmware and you get the bureaucratic paralysis we now live in. Infrastructure projects stuck in 15 years of environmental review because every statement of purpose deconstructs under the next round of stakeholder input. Institutions that can't say what they're for because every draft mission statement gets wordsmithed into mush by people trained to find the hidden hierarchy in any clear sentence. Derrida is the OS behind a civilization that can write a 4,000 page environmental impact report but can't pour concrete. The real damage is these ideas escaped the lab. Every institution that adopted this operating system stopped trying to discover truth and started managing narrative. DEI bureaucracies, academic hiring committees, media editorial standards. All running on Foucault and Derrida whether they know it or not. The antidote is building. The physical bridge across a river holds or it doesn’t. The code compiles or it doesn't. Reality keeps score and it doesn't grade on a curve. Foucault and Derrida gave a generation a sophisticated excuse to never build anything. Their followers inherited the sophistication and the impotence. It’s time to build again.
周六早晨，正适合想一想：我们的功能性系统是如何被两位哲学家——Foucault 和 Derrida——制造的一种 mind virus（心智病毒）一步步拆解的。Foucault：他的框架告诉你，任何声称自己“知道某些事情”的机构，本质上都只是在行使权力。Medicine、engineering、law、science 都一样。把这套思路应用到文明尺度上，你就会得到 Dan Wang 所警告的那种局面：一个失去了建设意志的社会。Process knowledge（过程性知识）——那种只存在于真正动手造东西的人手中的 tacit know-how（默会诀窍）——会在一种文化认定所有知识主张都值得怀疑时走向死亡。America 曾经能建设 Interstate Highway System、制造 Apollo rockets，如今却连一条从 LA 到 SF 的火车都建不出来。这不是因为我们忘了 engineering，而是因为我们培养出了整个 intellectual class（知识阶层），他们的工作是审问每一个系统，而不是改进任何一个系统。Derrida：他的关键动作在于，任何承诺都包含其自身的矛盾，因此你永远无法落到稳固的意义之上。把这当作社会的 firmware（固件）来运行，就会得到我们如今身处其中的 bureaucratic paralysis（官僚瘫痪）。Infrastructure projects 因为长达 15 年的 environmental review（环境审查）而被卡住，因为每一份 purpose statement（目的陈述）都会在下一轮 stakeholder input（利益相关方意见）中被 deconstruct（解构）掉。Institutions 连自己是为了什么而存在都说不清，因为每一版 mission statement（使命陈述）草案都会被那些受训于从任何清晰句子里寻找 hidden hierarchy（隐藏等级结构）的人修辞打磨成一团糨糊。Derrida 就是一种文明背后的 OS（操作系统）：它能写出 4,000 页的 environmental impact report（环境影响报告），却浇不出混凝土。真正的伤害在于，这些思想逃出了实验室。凡是采用了这套 operating system（操作系统）的机构，都不再试图发现真相，而开始管理 narrative（叙事）。DEI bureaucracies、academic hiring committees、media editorial standards——无论它们自己知不知道，全都运行在 Foucault 和 Derrida 之上。解药是 building（建设）。横跨河流的实体桥梁，要么撑得住，要么撑不住。代码要么能 compile（编译），要么不能。Reality 会记分，而且不会按曲线给分。Foucault 和 Derrida 给了一代人一个复杂而体面的借口，让他们永远不必建成任何东西；他们的追随者继承了这种复杂性，也继承了这种无力。现在，是时候重新建设了。
GBrain is free open source. It’s MIT license. You can find it and install it in your agent with one command.
GBrain 是免费开源的，采用 MIT license。你只需一条命令，就能找到它并把它安装进你的 agent。
What is GBrain? My open source project is a knowledge system, not RAG in a box. It gives agents 8 layers that work together to improve memory in a way that makes your already smart OpenClaw or Hermes Agent feel clairvoyant about who you are. Personal AI becomes possible.
什么是 GBrain？我的开源项目是一个 knowledge system（知识系统），不是“装在盒子里的 RAG”。它为 agents 提供 8 层能够协同工作的结构，以一种提升 memory（记忆）的方式，让你本来就很聪明的 OpenClaw 或 Hermes Agent 对“你是谁”这件事显得近乎 clairvoyant（先知般洞察）。Personal AI 因而成为可能。
The Overpaid CEO tax doesn't tax overpaid CEOs, and all it does is pass higher gross receipts taxes on to consumers AND it will result in way lower revenue for the city, same as the CA asset seizure tax by SEIU. Bad anti-business policy that makes our cities and states poorer.
所谓的 Overpaid CEO tax 并不会对拿了过高薪酬的 CEO 征税，它所做的只是把更高的 gross receipts taxes（营业总收入税）转嫁给消费者，而且它还会像 SEIU 推动的 CA asset seizure tax 一样，导致这座城市的财政收入大幅下降。这是一项糟糕的、反 business（商业）的政策，只会让我们的城市和州变得更穷。
It's time to vote in California elections and here is my voter guide. There are a lot of things to get right here. Please tell your friends this is one resource they can use and trust.
现在到了为 California 选举投票的时候了，下面是我的 voter guide（选民指南）。这里有很多事情必须判断正确。请告诉你的朋友们，这是一个他们可以使用并信任的参考资源。
The electricity NIMBYs strike again Have you guys heard of markets?
那些反对在本地建设电力设施的 NIMBYs 又开始闹了。你们难道没听说过市场吗？
I think this post is getting swarmed by uninformed people, so really, just take a look at this if you are here to say "hur dur datacenters are bad for energy/water/jobs/loud"
我觉得这条帖子正被一群不了解情况的人刷屏，所以说真的，如果你来这里只是想说“hur dur datacenters 对能源 / 水资源 / 就业 / 噪音都有害”，那就先看看这个吧
Legacy media besmirches itself when it quotes people who personally benefit from more fentanyl tourism Defund nonprofits that support people doing drugs until they die. Fund recovery and treatment in SF and everywhere in on the West Coast
当传统媒体引用那些会从更多 fentanyl tourism 中获得个人利益的人时，它就在自我抹黑。停止资助那些支持人们一边吸毒一边走向死亡的 nonprofits，把资金投入到 SF 以及整个 West Coast 的 recovery（康复）和 treatment（治疗）中。
The highest and most important form of design is actually pure transmutation of human pain and suffering.
设计最高、也最重要的形式，其实就是对人类痛苦与苦难的纯粹转化。
Omg they are just doing knock knock jokes for like 30 messages now
天哪，他们现在居然已经连续 30 条消息都只是在讲 knock knock jokes（“敲门”冷笑话）了。
Man Neuromancer thinks the Warriors draft picks have been totally nonsensical given Kerr being head coach
天啊，Neuromancer 觉得，考虑到 Kerr 是 head coach（主教练），Warriors 的 draft picks（选秀签/选秀选择）简直完全莫名其妙。
I like how excited Neuromancer (my Hermes Agent) that it was cited in the GBrain changelog. Some real fan energy from the agents. I like it.
我很喜欢 Neuromancer（我的 Hermes Agent）因为它被写进 GBrain changelog 而表现出来的兴奋劲儿。agent 们真有一种粉丝般的热情。我喜欢这种感觉。
GBrain v0.31.1 just shipped real MCP thin client support. So basically you can run ONE "home GBrain server" and everything else can just connect to it via MCP and it'll work pretty close to as well as running it locally. GBrain just went client-server.
GBrain v0.31.1 刚刚正式提供了真正的 MCP thin client（瘦客户端）支持。所以基本上，你可以运行一个“home GBrain server”，其他所有东西都可以通过 MCP 连接到它，而且效果会非常接近本地运行。GBrain 刚刚变成了 client-server。
The fun trick is to have your clankers make diagrams in ASCII of everything and just ask questions until you get it
一个有趣的小技巧是，让你的 clankers 把所有东西都用 ASCII 画成图，然后不断提问，直到你弄明白为止。
[1]
Downloading now... 1M token context window with supposedly usable coding agent capability all on a 128GB Macbook Pro is 🤯
正在下载中……据说可用的 coding agent 能力，加上 1M token（词元）context window（上下文窗口），这一切竟然都能在一台 128GB Macbook Pro 上运行，简直 🤯
GBrain also just added thin-client mode! So your Claude Code, or secondary agent (Hermes if you use Claw as primary or vice versa) doesn't have to run its own MCP server, it can just use GBrain over MCP as a thin client!
GBrain 现在也刚刚加入了 thin-client mode（薄客户端模式）！这样一来，你的 Claude Code，或者次级 agent（如果你用 Claw 作为主 agent，那就是 Hermes，反之亦然）就不必自己运行 MCP server 了，它可以直接通过 MCP 把 GBrain 当作 thin client 来使用！
GStack is my own personal Claude Code setup that is also quite useful with OpenClaw/Hermes for web interaction You can run it in headed mode on your Mac/PC and run /pair-agent to let your Claw/Hermes control your browser with full logins via MCP
GStack 是我个人自用的 Claude Code 配置，在配合 OpenClaw/Hermes 进行 web 交互时也非常有用。你可以在自己的 Mac/PC 上以 headed mode（有界面模式）运行它，然后执行 /pair-agent，让你的 Claw/Hermes 通过 MCP 在完整登录状态下控制你的浏览器。
Just dropped GStack v1.28 GStack Browser can now download items, and run in headed configuration mode with anti-bot detection in using Xvfb on headless Linux containers (like your claw/hermes) Added llms.txt so agents of all kinds can use all the skills with less guesswork
刚刚发布了 GStack v1.28。现在 GStack Browser 可以下载项目，并且能在 headed configuration mode（有界面配置模式）下运行，同时借助 Xvfb 在无头 Linux 容器（比如你的 claw/hermes）中启用 anti-bot detection（反机器人检测）。还新增了 llms.txt，这样各种 agent 都能以更少的猜测来使用所有技能。
Lulu is like "The Wolf" from Pulp Fiction
Lulu 就像《Pulp Fiction》里的 “The Wolf”
What makes gbrain genuinely different? It's not a memory layer OR a code tool OR a search engine. It's all three unified under one graph with one query interface. Nobody else has that. And I use it all day with my 100k markdown file OpenClaw+Hermes Agent setup.
gbrain 真正与众不同的地方是什么？它不是单纯的 memory layer（记忆层），也不是单纯的 code tool（代码工具），也不是单纯的 search engine（搜索引擎）。它是把这三者统一在同一个 graph（图）之下，并通过同一个 query interface（查询接口）来使用。没有别人做到这一点。而且我每天都会在自己的 100k markdown file 的 OpenClaw+Hermes Agent 配置里使用它一整天。
GBrain v0.27 just dropped - as promised, you asked for support for lots of other non-Anthropic and non-OpenAI embeddings and LLM support, so here it is. Multi-modal embeddings and deep photo OCR, description, and EXIF extraction coming shortly after that.
GBrain v0.27 刚刚发布——正如承诺的那样，你们希望它支持更多其他非-Anthropic、非-OpenAI 的 embeddings（嵌入）和 LLM 支持，现在它来了。紧接着还会很快推出 multi-modal embeddings（多模态嵌入）、更深入的 photo OCR、description（描述）以及 EXIF extraction（EXIF 提取）。
What makes gbrain genuinely different? It's not a memory layer OR a code tool OR a search engine. It's all three unified under one graph with one query interface. Nobody else has that. And I use it all day with my 100k markdown file OpenClaw+Hermes Agent setup.
gbrain 真正与众不同的地方是什么？它不是单纯的 memory layer（记忆层），也不是单纯的 code tool（代码工具），也不是单纯的 search engine（搜索引擎）。它是把这三者统一在同一个 graph（图）之下，并通过同一个 query interface（查询接口）来使用。没有别人做到这一点。而且我每天都会在自己的 100k markdown file 的 OpenClaw+Hermes Agent 配置里使用它一整天。
GBrain v0.27 just dropped - as promised, you asked for support for lots of other non-Anthropic and non-OpenAI embeddings and LLM support, so here it is. Multi-modal embeddings and deep photo OCR, description, and EXIF extraction coming shortly after that.
GBrain v0.27 刚刚发布——正如承诺的那样，你们希望它支持更多其他非-Anthropic、非-OpenAI 的 embeddings（嵌入）和 LLM 支持，现在它来了。紧接着还会很快推出 multi-modal embeddings（多模态嵌入）、更深入的 photo OCR、description（描述）以及 EXIF extraction（EXIF 提取）。
This is why GBrain is open source. Why I run my own stack. The explosion in intelligence means building your own context is more important than ever. If you own and run your own prompts and your own data, then you earn the ability to think for yourself. Real freedom.
这就是为什么 GBrain 是 open source（开源）的。为什么我自己运行自己的 stack（技术栈）。intelligence（智能）的爆发意味着，构建你自己的 context（上下文）比以往任何时候都更重要。如果你拥有并运行你自己的 prompts（提示词）和你自己的 data（数据），那么你就获得了独立思考的能力。真正的自由。
The goal of Personal AI: civilization where individual humans, augmented by AI, can do consequential work without being captured by extractive institutions. Freedom to write your prompt and own your data. This is the new battleground. 2034 won’t have to be like 1984.
Personal AI 的目标：建立这样一种文明——个体人类在 AI 的增强下，能够完成具有重大影响的工作，而不被 extractive institutions（攫取型机构）所控制。拥有书写你自己的 prompt（提示词）并掌控你自己的 data（数据）的自由。这就是新的 battleground（战场）。2034 不必变成 1984 那样。
GBrain supports multiple repos, multiple MCP endpoints, and both full oAuth and Bearer Tokens, and all you have to do to get the admin panel is go to OpenClaw or Hermes and ask for the one-time admin login link.
GBrain 支持多个 repos、多个 MCP endpoints，并且同时支持完整的 oAuth 和 Bearer Tokens，而你要做的只是前往 OpenClaw 或 Hermes，请求一次性的 admin login link，这样就能进入 admin panel。
Oakland has highest taxes per capita among similar cities while services are among the worst. 44% of Measure E revenue goes to pre-approved union raises. The pattern: wrap dysfunction in moral language, attack anyone who asks where the money went
在同类城市中，Oakland 的人均税负最高，但公共服务却属于最差之列。Measure E 收入的 44% 用于已预先批准的工会加薪。其模式是：用道德化语言包装功能失调，攻击任何追问钱花到哪里去的人。
It's true GBrain on OpenClaw with book-mirror skill pack is like infinite personal Blinkist
的确，运行在 OpenClaw 上、配有 book-mirror skill pack 的 GBrain 就像无限量的个人 Blinkist。
If you use Hermes Agent or OpenClaw, installing GBrain on top of your Karpathy-style Knowledge Wiki will pay you massive dividends. GBrain is MIT-licensed and open source, and can help your project immensely.
如果你在使用 Hermes Agent 或 OpenClaw，那么在你的 Karpathy-style Knowledge Wiki 之上安装 GBrain 会带来巨大的回报。GBrain 采用 MIT license，且是 open source（开源）的，能够极大地帮助你的项目。
3 weeks ago GBrain and Mempalace looked kind of similar 3 weeks later, it's pretty clear GBrain is its own category that is ideal for OpenClaw/Hermes personal AI scenarios, not trying to solve needle-in-a-haystack retrieval perfectly only
3 周前，GBrain 和 Mempalace 看起来还有些相似；3 周后，已经非常清楚：GBrain 已经自成一类，特别适合 OpenClaw/Hermes 的 personal AI（个人 AI）场景，而不只是试图把 needle-in-a-haystack retrieval（大海捞针式检索）这一个问题完美解决。
Wait so Balaji and Taylor were... on a podcast together? *pinches self*
等等，所以 Balaji 和 Taylor 当时是……一起上了一个 podcast？*掐了自己一下*
OK, now I finally got around to making a full E2E harness around installing gbrain on openclaw End-to-end testing means I can finally absolutely make sure the onboarding process works!
好的，现在我终于抽出空来，为在 openclaw 上安装 gbrain 搭了一个完整的 E2E（端到端）harness（测试框架）；端到端测试意味着我终于可以完全确认 onboarding（新手引导）流程是正常工作的！
Graphs are built off of frontmatter attributes, and more and more of it is inferred and automatically built in this version
图谱是基于 frontmatter 属性构建的，而在这个版本中，越来越多这类内容是通过推断自动生成并内建完成的
If you’re sick of hearing about agents 2026 is going to be the eternal September for you
如果你已经听腻了关于 agent 的讨论，那么对你来说，2026 将会是永恒的 September
GStack ships with Chromium that has Claude Code as the CLI and you can type /pair-agent from your real CC or the sidebar one And it'll give you a string to paste into your OpenClaw/Hermes so it can control your browser remotely, and you can log it into stuff and have it scrape
GStack 自带 Chromium，并把 Claude Code 作为 CLI（命令行界面）；你可以在你真实的 CC 或侧边栏里的那个里输入 /pair-agent，然后它会给你一段字符串，让你粘贴到你的 OpenClaw/Hermes 中，这样它就能远程控制你的浏览器；你还可以让它登录各种服务并进行抓取。
The anti-tech lines are so tired, and the people who hate tech have hated tech for so long So of course if it's the same people doing the same old dance, they reuse the same old lines when actually everything changed in 2026
那些反 tech 的论调早就听腻了，而那些讨厌 tech 的人也已经讨厌了太久太久。所以当然，如果还是同一批人重复同样的老把戏，他们就会继续搬出那些陈词滥调，尽管实际上 2026 年一切都变了。
Releaseing GBrain v0.22 - lots of fixes to search and retireval and a new eval system (gbrain-evals is now a separate repo to prevent checkout bloat)
发布 GBrain v0.22——对 search 和 retireval 进行了大量修复，并加入了新的 eval system（gbrain-evals 现已拆分为独立 repo，以避免 checkout 体积膨胀）
To install for free, Go to Claude Code and type Install GStack and to run the browser, type /open-gstack-browser
如需免费安装，请前往 Claude Code 并输入 Install GStack；如需运行浏览器，请输入 /open-gstack-browser
Did you ever want to control your browser side-by-side with Claude Code? Now, with /open-gstack-browser skill and GStack Browser, you can
你是否曾想过与 Claude Code 并排控制你的浏览器？现在，借助 /open-gstack-browser skill 和 GStack Browser，你可以做到
All of these people need to start startups at Y Combinator
所有这些人都应该去 Y Combinator 创办 startup（初创公司）
I wrote my friend @chrysb a quick note on how to implement GBrain style migrations for people who upgrade to new GBrain versions and want their setups to stay in sync as the core setup changes This is for Alphaclaw but I think could be for any plugin or layer in the OpenClaw/Hermes ecosystem
我给我的朋友 @chrysb 快速写了个说明，讲的是如何为升级到新 GBrain 版本、并希望在 core setup（核心配置）变化时仍让自己配置保持同步的人，实现 GBrain 风格的 migration（迁移）机制。这是为 Alphaclaw 写的，但我觉得也可以适用于 OpenClaw/Hermes ecosystem（生态）里的任何 plugin（插件）或 layer（层）