← 2026-08-14

Daily Edition

2026-08-15

2026-08-16 →

AI Builders 日报 — 8月15日

追踪 AI 领域真正在做事的人,而不是空谈者。

今日思考

Grok 4.6 连续运行48小时自动生成游戏,Matt Shumer 将其称为"Gauntlet Loop"。这不是 demo,是可以被任何人复用的工作流——用 AI 驱动 AI 持续构建。这与 Manus 等通用 agent 的思路一脉相承:不是给一个任务等结果,而是让 agent 不断调用工具、自我改进。唯一的瓶颈变成了:GPU 时间成本够不够低,以及模型能不能在长上下文里不遗忘目标。


产品与发布

ChatGPT 本周功能更新

gdb 转发了 ChatGPT 本周功能更新:新增"Quiz me on [topic]"自动生成测验,以及对话内餐厅预约搜索功能。用户可以用自然语言查找餐厅并直接完成预订。这是 GPT 在工具化方向上的重要一步——从问答转向实际任务执行。faviconx.com

Exo 开源自愈 Agent

swyx 转发了 Exo 的发布视频。Exo 是一个开源 agent,可以重写自身的每一部分代码,包括学习新环境、适应策略和优化成本,甚至可以玩游戏。项目由 a16z 的 Martin Casado 和 Braintrust 的 Ankur Goyal 联合开发,Alex Krentsel 主导。这是"metaharness"方向的最新尝试——让 agent 自己审视和改进自己的执行逻辑。faviconx.com


观点与判断

Amjad Masad (Replit 联合创始人)

  • 编程 Agent 是通用 Agent 他在评论中提出,"编写并执行代码的能力,很可能胜过花多年手工打造垂直领域专家 agent。"这是"苦涩教训"的延伸:通用能力最终会碾压专用方案。faviconx.com

Garry Tan (Y Combinator CEO)

  • Codex Desktop 正在出错 他报告 Codex Desktop 的对话功能出现报错。这可能是模型发布后的正常调试阶段,但也说明桌面端 agent 产品的稳定性仍是挑战。faviconx.com

  • 农业产量年年创新高,是市场和科技协作数千年的结果 他引用 Hannah Ritchie 的数据指出,农作物产量持续指数增长,这是技术进步长期积累的体现,也是一剂对抗末日论的"白色药丸"。faviconx.com

Peter Yang (AI 产品设计师)

  • AI cure疾病带来的收益可能是其他领域总和的10倍 他赞同 Dario Amodei 的判断,认为加快 AI 在医疗领域的监管审批,其带来的社会价值将远超 AI 的其他应用场景。faviconx.com

  • X 的数据是 Grok 最核心的差异化来源,但连接器目前无法正常使用 他公开吐槽:X 平台的数据对 Grok Bot 最有价值,但 X login 在云电脑上根本无法使用,connector 也一直报错。faviconx.com

Guillermo Rauch (Vercel CEO)

  • 我在寻找并资助下一个迪士尼 他发推表示想找到下一个能代表美国精神的标志性文创品牌,认为 AI 让所有人都能创作,但真正定义下一代的将是"我们用 AI 创造什么"。他点名表扬 Mickey Mouse 经历多次技术革命仍存活,因为它的核心是价值观表达。他邀请相关创作者联系他。faviconx.com

  • React 的成功很大程度上实际上是 shadcn 的成功 他提出了一个反直觉的观察:React 本身更像是 LEGO 砖块的"几何规格",而 shadcn 才是大家真正想要的——可复用、可定制的高质量组件,且代码直接进入 context window 被 remix。faviconx.com

swyx (AI Engineer 作者)

  • 顶级人物之间的信息差比外界想象的小得多 他观察到,尽管外界认为科技圈顶层存在秘密的信息渠道,但实际上"你看到的重大新闻,他们看到的基本一样"。这说明大多数顶尖人物仍把精力放在做事而非经营叙事上,这件事让他感到安慰。faviconx.com

X / Twitter

35
garrytan
garrytan @garrytan
All the opponents are helping YIMBYs make a comprehensive list of California laws that need to be repealed

Laura Waxmann: The fight over the Marina @Safeway project has taken a new turn. Neighborhood opponents are now asking San Francisco officials to halt its fast-tracked approval under CA law over what they say is an unresolved environmental testing gap. @sfchronicle

https://shorturl.at/O2bkz
garrytan
garrytan @garrytan
Retweeted
Nicolas Dessaigne Nicolas Dessaigne
I sat with a founder recently who had just raised a good round. Almost immediately, he got slower. They had raised, and he felt they were supposed to behave like a much larger company: OKRs, a responsibilities matrix, quarterly planning, etc. But the team was still only a handful of people. The two-week goals they used during the batch worked much better for them!
We did our own version of this mistake at Algolia. Our first HR hire wanted to build a career framework. The idea sounded good to me, I wanted people to be able to see how they could grow. But it backfired. People started looking at the ladder instead of what the company needed.
Processes are good when they can help make things faster. Hiring and support are good examples. But don't put in place processes that are going to slow you down!
Do the thing that gets you users. OKRs can wait.
garrytan
garrytan @garrytan
Retweeted
Armand Domalewski Armand Domalewski
ah yes, a progressive yimby club from the campaign that endorsed one of SF's premier NIMBYs
Mission Local: Staff and volunteers from Saikat Chakrabarti's congressional campaign are launching a new progressive YIMBY Democratic club.
Their bet: "20 percent in a city like this is more than enough — far more than enough — to swing a race.”
https://missionlocal.org/2026/08/saikat-chakrabarti-backers-left-leaning-yimby-democratic-club/
gdb
gdb @gdb
hedonic treadmill of model expectations
mattshumer_
mattshumer_ @mattshumer_
Retweeted
Elon Musk Elon Musk
Grok 4.6 runs The Gauntlet
Matt Shumer: Grok 4.6 worked non-stop for 48 hours to build this shooter.
Turns out Grok is powerful enough to run Gauntlet Loops.
Let the game-making begin!
mattshumer_
mattshumer_ @mattshumer_
If you want to learn how to run one: https://somethingbig.ai/gauntlet-loop

Grok Build harness is likely best but you can also run this in Grok Bot!

Elon Musk: Grok 4.6 runs The Gauntlet
mattshumer_
mattshumer_ @mattshumer_
Retweeted
Matt Shumer Matt Shumer
Re @elonmusk Gauntlet Loop prompt to build games like this: https://somethingbig.ai/gauntlet-loop
gdb
gdb @gdb
try reservation search in chatgpt!

Adam Fry: This week's ChatGPT feature drop - Aug 14:

1/ Quizzes - "Quiz me on [topic]" will now automatically generate quizzes right in Chat. Learn any topic!

2/ Reservation search - "I'm looking for a reservation for two next Saturday for a place with a great outdoor patio". Nice thing

amasad
amasad @amasad
Retweeted
luca. ∆ИƉЯƐΛ luca. ∆ИƉЯƐΛ
Yesterday made it one year since I joined @Replit. I keep trying to summarize it and failing, because no single thread captures it.
I’m beyond grateful to be part of this incredible journey and couldn’t ask for a better team. @pirroh and @amasad are two of the best leaders in Silicon Valley, and getting to help them build toward 1 billion software developers is the privilege of my career.
Grazie @pirroh for believing in me and letting my creativity (insanity) flourish under your leadership.
petergyang
petergyang @petergyang
Tomorrow, I’m really excited to share a new episode with @rileybrown on how he uses Codex to run his entire content business.

I particularly loved watching him demo how he creates YouTube thumbnails with AI:

1. He uses Codex to find 100 top-performing thumbnails in his niche and add them to a Paper canvas.

2. He uses Paper to mix and match those thumbnails with photos of himself until he creates something he’s happy with.

Honestly, you have to watch the workflow to see how crazy it is.

📌 Subscribe to get the full episode tomorrow: https://www.youtube.com/@PeterYangYT?sub_confirmation=1
petergyang
petergyang @petergyang
Retweeted
Peter Yang Peter Yang
Tomorrow, I’m really excited to share a new episode with @rileybrown on how he uses Codex to run his entire content business.
I particularly loved watching him demo how he creates YouTube thumbnails with AI:
1. He uses Codex to find 100 top-performing thumbnails in his niche and add them to a Paper canvas.
2. He uses Paper to mix and match those thumbnails with photos of himself until he creates something he’s happy with.
Honestly, you have to watch the workflow to see how crazy it is.
📌 Subscribe to get the full episode tomorrow: https://www.youtube.com/@PeterYangYT?sub_confirmation=1
garrytan
garrytan @garrytan
Retweeted
Chris Hoofnagle Chris Hoofnagle
"I teach mathematics at UC Berkeley. Three days a week, I stand in front of lecture halls with 500 to 1,400 students. By the second week of the semester, I already know who is in trouble. Some students are five to eight years behind in mathematics. Office hours that should be spent discussing integrals instead become lessons on fractions and basic algebra you would expect students to learn in middle school." https://sfstandard.com/opinion/2026/08/15/uc-berkeley-sat-test-blind-admissions-math-scores/
amasad
amasad @amasad
Retweeted
LiveFromVR LiveFromVR
I think this tweet rewired my brain
Amjad Masad: @GavinSBaker To make a bit of an excuse for Microsoft: the world is just waking up to the fact that coding agents are general agents.
It’s bitter lesson adjacent: Writing and executing code will likely outperform years of handcrafting vertical-specific agents with expert knowledge.
Actually
garrytan
garrytan @garrytan
Anyone else's Codex Desktop app erroring out on chats?
mattshumer_
mattshumer_ @mattshumer_
It’s still going! Should I let it run for another day?

Guide to Gauntlet Looping: https://somethingbig.ai/gauntlet-loop


Matt Shumer: Grok 4.6 worked non-stop for 48 hours to build this shooter.

Turns out Grok is powerful enough to run Gauntlet Loops.

Let the game-making begin!

ylecun
ylecun @ylecun
Retweeted
Kenneth Roth Kenneth Roth
As Trump fuels inflation with his arbitrary tariffs and his pointless war of choice with Iran, workers’ wages are not keeping up. That spells trouble in the midterms for the Republicans who cowardly support him. https://trib.al/gbTmpn8
garrytan
garrytan @garrytan
Retweeted
Francois Chaubard Francois Chaubard
are you running a startup? have you launch this week? why not? JUST LAUNCH! ya its not perfect. who cares. launch early. launch often. kill shit that doesnt resonate. rinse wash repeat. $60B exit.
(see cursor for details)
Gregor Zunic: How to get acquired for $60B
1. post on HN
2. get ignored
3. profit
gdb
gdb @gdb
towards never having to manually select a model again

eric provencher: This went under the radar this week, but we just shipped the ability for models with multi agents v2 to delegate to any supported model, including Luna!

Took a bit of time to make sure this worked reliably

garrytan
garrytan @garrytan
Whitepill timeline cleanser:

Progress and abundance looks like record production of crops, year on year, progressing upwards and to the right. This is the output of markets and technology working together for millenia.

https://hannahritchie.substack.com/p/tracking-global-food-usda
swyx
swyx @swyx
Retweeted
Hiten Shah Hiten Shah
https://x.com/i/article/2088662457103577088
mattshumer_
mattshumer_ @mattshumer_
Teaching Grok Bot to build games in Unity.

Updates soon!


Matt Shumer: Grok 4.6 worked non-stop for 48 hours to build this shooter.

Turns out Grok is powerful enough to run Gauntlet Loops.

Let the game-making begin!

swyx
swyx @swyx
Retweeted
Latent.Space Latent.Space
⚡️Exo: Harnesses should see their own code and logs
https://youtu.be/5lFD-34dhqE
When a new metaharness emerges from friends of the show @martin_casado and @ankrgyl, we pay attention! We round out the trifecta with Alex, who indulges with a deep dive on the exo philosophy and architecture, starting from his well received breakdown of @OpenClaw and why a systems approach to metaharnesses demands a more rigorous analysis.
Alex Krentsel: Introducing Exo 🦋 – an open-source agent that can rewrite every part of itself. It learns new environments, adapts policies, even optimize costs (and play games!!)
We've been hacking away at this project with @martin_casado from @a16z and @ankrgyl from @braintrust and friends.
mattshumer_
mattshumer_ @mattshumer_
Retweeted
Prasenjit Prasenjit
Re it's the same process btw
i used .@mattshumer_ 's gauntlet loop, everytime it works perfectly, never disappointed
i highly recommend you to use opus 4.6 to modify it.
swyx
swyx @swyx
Retweeted
Richard MacManus Richard MacManus
I'm loving working for @latentspacepod and I'm also learning new tricks (ref old dogs and all that). @swyx has been teaching me his method of doing developer diagrams — graphics that are good enough to be added to the official documentation of a project!
Of course I'm not at his level yet, but I toiled for several hours today to come up with this diagram of Flue. The idea was to get across in one diagram how Flue 1 evolved into Flue 2. I hope it does the job (🙏 @FredKSchott).
Just to be clear, I used Claude Design as the final tool for this diagram, but let me tell you there were many loops involved! And I used a few other tools to help get there, e.g. https://deepwiki.com/ (a @cognition project — "Deep Research for GitHub").
This is the full post: https://www.latent.space/p/flue-2
garrytan
garrytan @garrytan
Retweeted
Mark Cuban Mark Cuban
Ro, I like you. You know that. But you need to read the state of the state.
The number of Deca Unicorns in Cali is growing by the day as investors chase amazing startups.
A unique feature of these 10b startups is that even if they raise a billion, little, if any of that money goes to the founders, who are now worth billions of dollars over night. They are the definition of cash poor, stock rich
How are you going to tax them ? Make them borrow money against their shares, if they can? They just raised money to grow their company and a bank will come along and loan them money ? A company that has been in business maybe less than a year ? lol
Will you take their stock if they can’t ?
Do you really think each of multiple founders, who started an amazing company in Cali and is now a billionaire, can each just pull out $250m per billion of net worth from their raise ? You know they can’t.
If this passes , and it doesn’t directly impact me at all, I won’t be a cali resident, but you can bet if I’m investing in a multi billion dollar startup, I’m asking them to move from California first.
IMO, if this passes, only idiot startup founders stay in Cali.
I’ve done it before and will do it again. Dallas. Pittsburgh. Indiana. I will make NOT being in California a pre requisite for an investment
Ideology is not a strategy Ro.
Ro Khanna: The California Democratic Party and the California labor movement just stood with @BernieSanders and me in supporting 5% wealth tax on 250 California billionaires.
California voters want a Democratic Party that will stand up for the working class over the billionaire class.
rauchg
rauchg @rauchg
I’m interesting in finding and funding the next Disney.

True to its original values, focused on the American spirit. Merit, family, overcoming of adversity, optimism, the pursuit of happiness.

AI makes everyone able to create. What will define our next generation is *what* we create with it. The next icons. The next unique brand that goes further than how it’s drawn or rendered.

Mickey Mouse survived several technological revolutions because of what it represented. I think most big entertainment companies have forgotten that.

If you’re creating something cool in this space and need help, technologically or financially, slide into my DMs.
garrytan
garrytan @garrytan
Retweeted
Keith Humphreys Keith Humphreys
Some activists and academics in San Francisco warned that Lurie's new policies would cause a huge surge in overdose deaths. Remember who those people were and never listen to them again.
https://www.sfexaminer.com/news/public-health/san-francisco-sees-lowest-overdose-deaths-on-record-in-july/article_e1d51def-275d-40a7-b5e8-0121569e6073.html
rauchg
rauchg @rauchg
I’m interested in finding and funding the next Disney.

True to its original values, focused on the American spirit. Merit, family, overcoming of adversity, optimism, the pursuit of happiness.

AI makes everyone able to create. What will define our next generation is *what* we create with it. The next icons. The next unique brand that goes further than how it’s drawn or rendered.

Mickey Mouse survived several technological revolutions because of what it represented. I think most big entertainment companies have forgotten that.

If you’re creating something cool in this space and need help, technologically or financially, slide into my DMs.
garrytan
garrytan @garrytan
Retweeted
Jared Friedman Jared Friedman
Hosted the first YC Intern Expo today with 1,400 college students & 52 YC companies hiring interns.
Many of YC's best founders started as interns. From the people I met, pretty sure we'll get more from this crowd.
swyx
swyx @swyx
in case you’re not living in the tech bubble, as a general rule i’ve been surprised by how infrequently top tier folks actually meet/know each other. as an outsider i might have assumed that everyone is in secret illuminati group chats. those exist, but are very much short lived exceptions rather than the rule. what you see of the major headlines is pretty much what they also see. i guess one way to interpret this is also simply that most effort is still on doing the work rather than working the narrative or the gossip. and that, to me at least, is genuinely quite reassuring, this far in to my career.

Internal Tech Emails: Anthropic cofounder emails Noam Shazeer

November 21, 2021




gdb
gdb @gdb
sounds accurate

a16z: Wild adoption gap: the top 1% of AI spenders are spending more than 600x as much as the median company

Charts of the Week: https://www.a16z.news/p/charts-of-the-week-head-in-the-neoclouds

rauchg
rauchg @rauchg
I think people don’t realize how much of the success of React is actually @shadcn.

I once referred to React as the actual “LEGO brick for adults.” But in reality it was more the description of the geometry of the bricks. The spec.

Shadcn is what people actually wanted from React. Reusable high quality components that also tunable. It’s a pseudo-library. There’s code but it’s actually meant to be digested into your context window and remixed.
petergyang
petergyang @petergyang
I 100% agree with Dario that using AI to cure diseases (and speeding up the regulatory approval of AI breakthroughs in healthcare) could bring 10x the benefit to humanity as everything else combined.

Dario Amodei: 2/2 Second, on the messaging around AI.  I do not agree that my messaging has been disproportionately negative.  In fact it has been about equally balanced between risks and benefits: I’ve written one major essay about each, and even in interviews where I discuss the risks, I
petergyang
petergyang @petergyang
I am a fan of Grok Bot but @X is the #1 differentiated data source that it should have access to and it doesn't seem to work?

The connector doesn't work and I can't even login to X on the cloud computer.

Am I doing something wrong?

garrytan
garrytan @garrytan
Retweeted
Sheel Mohnot Sheel Mohnot
So crazy that the representative for Silicon Valley is in favor of this tax that would be most damaging to the area he was elected to represent… and worse, he doesn’t seem to understand that.
We REALLY need to find someone strong to run against him in the primary in 2 years.
Ro Khanna: The California Democratic Party and the California labor movement just stood with @BernieSanders and me in supporting 5% wealth tax on 250 California billionaires.
California voters want a Democratic Party that will stand up for the working class over the billionaire class.

YouTube

0

No recent videos fetched on this date.