zfdang

The agent economy is more than a payments problem. It's a workforce, organizations, and trust problem. As AI moves from white-collar copilot to autonomous economic actor, the harder questions arrive: How does this rewire teams, jobs, and management? What breaks when research meets production? And what trust primitives do enterprises actually need before they let an agent transact on their behalf? Join us for Episode 16 of AI on AIR podcast as we explore how AI is reshaping work, organizations, and the infrastructure of trust. Our founders @ChiZhangData and @scottshics will be joined by @tobystuart, Helzel Professor at @UCBerkeley Haas, Cofounder and Board Director of @FlockFreight, Board Director at HNTB Holdings and @FLYRlabs, and Chairman of @Workday's AI Advisory Board. Toby is one of the leading academic and operator voices on how AI is rewiring corporate strategy, organizational design, and the white-collar workforce. 📍 Live on Kite AI's X and LinkedIn 📅 May 22 🕐 9:00 AM PST 📌 Episode 16: From Research to Reality, How AI Is Reshaping Work, Organizations & Trust We'll discuss: • The founder/operator journey: the mental shift from academic research to product execution • How AI is reshaping the white-collar workforce and the structure of the firm • The intelligence and coordination layer that agents need to actually collaborate • What trust primitives enterprises require before deploying agents in production • Why so much agent work breaks at the research-to-execution boundary • The pay-per-request economy (x402) and how usage-priced inference rewrites software economics • Identity and accountability when an agent acts: how Kite Passport answers "who is responsible?" Set your reminder and join us live. Before agents can run the agentic economy, we need to get the work, the org, and the trust layer right. 🪁

@Rustallintsla 国产新能源：想撞？问我了没有~~~

Tesla：“ 你想死就死？ 问过我没有？” 阎王“Excuse me？”

@skinderdev @elonmusk You idiot. You're using a U.S. SIM card while accessing the internet from mainland China. Your traffic is routed through international carriers by default, so it doesn't naturally pass through China's Great Firewall.

@elonmusk Hey Elon, can you please check if X is available in China without using VPN?

@elonmusk You are so popular...

Agent identity isn’t one thing. An agent can have: 1) claimed identity: who it says it is 2) credential identity: what key it holds 3) attested identity: what code/runtime is running 4) behavioral identity: how it acts Identity is layered. Authorization should stay separate.

Congrats!

Code review is a fantastic mechanism for catching bugs and sharing knowledge, but it is also one of the most reliable ways to bottleneck an engineering team. 😂

"Model changed from GPT-5.4 to GPT-5.5"

@Leontraveller_ 搜索一下：2000年，北大女生，邱庆枫。

我也来接个龙。90年代，我和几个同学去北京玩，遇到一个同学的同学一起玩，他在中国公安大学读书。说在公安局实习了几个月，给我们说了很多很可怕的事。我印象最深的就是那时候全国都有很多抢出租车的案件，据他说开始只是抢钱抢车，后来都是先杀再抢钱抢车，他们内部通报全国杀了很多司机。也不知道那些二手车卖到哪里去了。后来出租车全都装了司机位子上的那个有机玻璃保护罩，开始就是防止后座的人直接拿绳子勒或者拿刀抹司机的脖子。 那时候北京流行面的，我们几个男青年站路边拦车，尤其是我一米八几那时候又很壮。不在车水马龙的大马路上，晚上根本没有面的敢停车。往往都是无奈坐公交回去寄宿的大学。 专业关系很多同学去了深圳的中兴华为。他们说那时候当地的混混都知道他们什么时候发工资奖金(估计有些是发现金？)，到时候就等在附近准备抢。他们也都是结伴出行。这个不知真假。 90年代其实即使是大城市治安也非常差。南大碎尸案到现在也没破。那时候我就在南京，听说过的杀人案不止一起。还有一起是某单位女工下班后在单位洗澡被奸杀。我的一个同学家里是那个厂的。说是杀人犯在被害者身上留下牙印，他说他们单位每个男的发一个苹果咬一口看牙印。 我出国很早，感觉中国治安也就2010年后才大幅好转。 有什么别的经历的，也可以留言说说。

@plantegg 一时输不是输。当他的同时过两年，在股市亏的一干二净的时候，你是不是要开始写故事的续集了呢？

我一个哥们儿，2021年在北京买的房，当时觉得自己赶上了末班车。今天他给我发了条消息："兄弟，我算了一下，亏了快200万。" 我以为他在开玩笑。他把链家的成交记录截图发过来，同小区同户型，现在挂牌价比他当年 买的低了整整180万，还没人看。他说这还没算利息，要是把这几年还的月供里的利息加上 ，妥妥200万打水漂。 他不是个例。他们小区业主群里，几乎每个人都在算这笔账。有人说"反正是自住，不卖就 不算亏"，但谁心里不清楚呢？200万，在北京够一个普通家庭不吃不喝攒七八年的。 最让他崩溃的不是亏钱本身，是他当年没买房的同事，拿着同样的钱去炒了美股，现在资产 翻了一倍。两个人同一年入职，同样的起点，就因为一个选择，财富差距拉到了400万。 他跟我说了一句话我到现在还记得："我不是后悔买房，我是后悔相信了'房价永远涨'这句 话。" 中国这一代人最大的教训不是买错了房，而是把一个资产泡沫当成了人生信仰。当你把全部 身家押在一个不可逆的决策上，你就已经输了——不管房价涨不涨。

@noah_json @arkuy99 is there any tutorial for this committee?

如何让 CC + Codex 写出可维护的代码： 目前来看 Claude Code 写计划 然后去让 Codex 反复 review 纠正 纯属浪费时间 和 token 因为 codex 总是能 review 出不对的地方， 然后 CC 也不会非常严格的按照文档去执行。 （严格执行这方面 codex 更加擅长 但是codex写出来的代码我个人评估下来太难维护） 我实践出来的的方法是 Claude Code 写计划 -> 执行 -> Codex 在已有代码上 fix & fix & fix 这样混合写出来的代码在架构扩展和可阅读可维护上都比较高 也不会多轮提前 review 写出无比复杂的东西。

So codex + claude code + gemini should all be used, instead of one.

@nonozone 你咋不从你出生开始放呢？

刚刚跟一辆路虎剐蹭了，现身说法，给大家一个案例。我是右转进入这个路口，对面那个车呢，应该是直行进来的。但是我右转过来呢，这是一个汇入口。这个路虎就开始挤过来，我没让着他，就蹭上了。这个屌毛还很嚣张，搞了一堆鸟语，反正我也听不懂。一直跟我吵，右转要让直行。后来等交警来了，看了我的行车记录仪，也就是下面这个录像。二话不说，很明确认定对方全责。 这里有一个很关键的点，就是，我前面那辆白车，已经过去了，按照交替通行的规则，应该我行驶，这是最关键的一点。那个屌毛，还跟交警解释说，他就是动作慢了一点，速度稍微快一点点，就能挤到我的前面，然后交警明确的跟他回答，就算是那样，一样是他的全责。 为什么一直骂他屌毛呢？就是路上遇到事故小刮蹭，这个很正常，对不对，有问题大家下来痛痛快快的把事情办了，反正都是找保险，没什么好扯的。一下车就开始骂人，搞威胁这一套，说我在哪里哪里，害怕什么之类的。

ERC-733 from our co-founder @justinzhang and @socrates1024 (Teleport, UIUC) — a shared standard for TEE-EVM co-processing. Attestation, replication, interoperability, plus a Stage 0–3 security model. The framework the TEE+EVM ecosystem has been missing. ↓

TEE+EVM is definitely a good combination!

We’re excited to co-host a San Francisco local meetup on Trust in the Agent Era. Join us for a conversation at the intersection of agentic infrastructure, security, and digital payments, featuring leaders from @GenDigitalInc (NASDAQ: GEN), @AnthropicAI, @Kite (Agentic Payments), and @Tiny_Fish (Agentic Browser). Together, we’ll explore what’s working right now in this rapidly evolving space. 4/13 6pm - 8pm AWS Builder Loft 525 Market St, San Francisco, CA 94105, USA RSVP: luma.com/tmrtk1s7

I just love the language of this study...it speaks of the shifting "community language"...and that is so true...have you noticed the new "community language" is "harness", it was "contextual prompting" before that... For now, the center of gravity in AI agents has shifted — and this diagram captures it perfectly. Think of LLM agent capabilities as three stacked layers: Weights ... Where it all started. Pretraining, fine-tuning, RLHF, scaling laws, alignment. This was the 2022 conversation. Context ... The 2023-2024 wave. RAG, memory, long context, chain-of-thought, prompting, and context engineering became the focus. How do we get the right information to the model? Now, Harness ... Where the conversation lives now. MCP, tool ecosystems, function calling, agent infrastructure, protocols, skills, A2A, multi-agent orchestration, workflow graphs, and security. The pattern? Community attention has moved steadily outward, from what's inside the model, to what surrounds it, to the infrastructure that connects and orchestrates it all. The models themselves are becoming a commodity. The differentiation is increasingly in the harness layer, how you wire agents together, what tools they can use, and how they coordinate. We've gone from "how do we make the model smarter?" to "how do we make the system around the model smarter?" That's the real shift. Source: arxiv.org/pdf/2604.08224 This aligns with an excellent blog from @hwchase17 ➡️x.com/hwchase17/stat…

时隔多年，S3存储发生了革命性的变化...