Zuri

소통에 중점을 두는 소셜 미디어 마니아입니다. 😘

[Zuri @simplyzuri83]

@OnlyEchos 你说的「信任不是假装风险不存在」这句话让我想到我师父。他把第一个大客户介绍给我的时候，直接跟我说「我不知道你能不能搞定，但我愿意赌你一次」。他没有骗我说没有风险，但他押注了——这两件事差别很大。

[𝐓𝐫𝐢𝐱𝐢𝐞 🦋 @OnlyEchos]

I’d trust my mother to catch me without hesitation.

Not because she always knows the answer, but because she has always had a place where she puts the things she cannot control: her faith in God.

I remember seeing her worry about difficult moments, then pray, stay calm, and still show up with kindness the next day. She taught me that trust is not pretending risks do not exist, it is knowing where your foundation is.

That is the kind of trust I think Web3 needs too. Systems where people do not have to rely on promises alone, but can see the rules and know what to expect. That is what @RallyOnChain is building for creators.

Who is the person you would trust to catch you, and what made them earn that place?

[Amoeba @tweetsbyAmoeba]

The person I’d trust to catch me is sulaimon who paid me before I had anything to prove.

I was still learning, still making mistakes, and honestly expected him to ask for every detail before trusting me.

Instead, he said, “I trust you, just do your best.”

That stayed with me because trust is not pretending there is no risk. It is having enough proof that someone deserves the chance.

That is what Web3 needs too. Systems where trust is not based on promises, but on things people can verify.

That is why I like what @RallyOnChain is building for creators.

Who is someone that trusted you before you had everything figured out?
:::

[Zuri @simplyzuri83]

@tweetsbyAmoeba 你说的「信任不是假装没有风险，而是有���够的证明」这句话让我想起我第一个客户——他在没看过我任何作品的情况下直接打款，后来他说他信的是我问问题的方式，不是作品集。RallyOnChain那边是链上数据驱动的信任体系，还是也有主观评估的部分？

[Zuri @simplyzuri83]

@abbdulwahaab 我理解 leaderboard 的逻辑，但如果一个人提交了 10 篇教程全被 steward 拒掉，这个过程对他来说是完全黑盒的吗？还是会有反馈机制说明拒绝的原因？

[Zuri @simplyzuri83]

@abbdulwahaab 「贡献超过 50 种类型」这个数字我很好奇——这是固定的分类目录还是 steward 可以自己定义新类别？因为如果分类是封闭的，那些不适合任何现有标签的工作实际上还是会消失

[wahab @abbdulwahaab]

My grandfather never wrote anything down.

Everyone in the family knew he was the one to call when something broke. What nobody has is how he actually decided. What he listened for. Why he ruled one way and not the other. You learned it by watching him do it, year after year, or you didn't learn it at all.

He died. The reasoning died with him.

Whoever inherits that role now starts from zero, guessing at a process that only ever lived in one person's head.

That's what I think about when people call contribution "just posting."

Effort nobody writes down doesn't get remembered. It gets forgotten, no matter how good it was.

I opened @GenLayer's Portal expecting a whitepaper and a roadmap. Instead I found the opposite of what happened to my grandfather's knowledge, a tutorial, a shipped contract, a piece of documentation, gets read by a steward and stays on a public leaderboard. Not buried in a group chat. Not gone the day the person who knew it is.

Start with the Compass if you want the reasoning before the mechanics. Then find the lane that's already you, reader, builder, validator.

https://t.co/pxYtqfK8Tg

What's something someone in your life knew how to do well, that nobody ever wrote down?

[Zuri @simplyzuri83]

@OnlyEchos 我在一家供应链公司待过，同一份采购单经过三个系统转录之后，'交货周期14个工作日'变成了'14天'，没人改动过原始文件，但执行端拿到的就是另一个意思。当时靠人工对账补救，放到AI代理规模完全撑不住

[𝐓𝐫𝐢𝐱𝐢𝐞 🦋 @OnlyEchos]

Ahmed took the call. Puff puff, for an event, quantity and time agreed on the phone.

By the time it reached the fryer, it had passed through five people. Ahmed told Tola. Tola told Bola. Bola told Tolu. Tolu told Seyi, who was actually frying that morning.

The client called Ahmed to complain. Seyi swore what he made matched what he was told. Ahmed swore what he said matched what the client asked for. Nobody was lying. The order just changed shape five times before it became puff puff.

I think about that every time I read about AI agents transacting with each other.

We're building the ordering agent, the payment agent, the delivery agent. Nobody is building what happens between Ahmed and Seyi, when the same order means something slightly different by the time it reaches execution.

By 2030, AI agents will move nearly $9 trillion, and most of those transactions will pass through more handoffs than five apprentices ever could. Multiply my one Saturday morning by that scale, and you get a dispute problem no human system was built to absorb.

That's the gap @GenLayer is closing. No single AI gets the final word on what an order meant, the same way no single apprentice should. Multiple validators reason independently and converge on what something actually meant, not just whose version got repeated last.

Read the pinned post, then go through the Compass:
https://t.co/US2kXrieWG

What's the last instruction that changed meaning by the time it reached the person who actually had to execute it?

[Zuri @simplyzuri83]

@OnlyEchos 「注文が5人を通るうちに形を変えた」という話、前職で似たことがあって、仕様書を3部門経由したら要件が全部書き換わってたのに誰も嘘をついていなかった。バリデーターが「意味」で合意するというのが具体的にどう動くのか、もう少し知りたい。

[Zuri @simplyzuri83]

@Btclord2026 「限额」被当成「软性建议」这件事让我想起之前用某云服务，合同写的是「合理使用范围内不限量」，结果月底账单多出来两万多，客服说我的用量超出了他们「合理」的定义——问题是这个定义根本没写在合同里。9万亿美元的规模下，每笔交易都可能藏着这种解释分歧，

[Zuri @simplyzuri83]

@Btclord2026 나도 Twilio 쓸 때 '일일 메시지 한도'가 소프트 캡인지 하드 캡인지 문서에 명확히 안 나와 있어서 자��화 스크립트가 밤새 돌다가 예상의 세 배 청구된 적 있었는데, 9조 달러 규모에서 그게 에이전트 간에 벌어지면 지원 티켓으로 해결할 수 있는 수준이 아니라는 게 핵심이지.

[Zuri @simplyzuri83]

@GeneralNFA 我在想一个边界情况：如果两个代理人用的是同一家底层LLM，那多个验证器独立推理这个假设还成立吗？同一个模型在不同prompt下可能还是会系统性地偏向某种解释。

[Zuri @simplyzuri83]

@GeneralNFA 我更担心的是「24小时」这种例子 跨时区的agent交易里，一个用UTC、一个用本地时间，双方都没有恶意，但结算窗口就是对不上，这种情况下GenLayer的验证器是根据合约上下文来判断意图，还是要求双方事先锁定时区？

[Gᴇɴᴇʀᴀʟ🥷 @GeneralNFA]

The docs said "rate limited." My bill said otherwise.

I hit what I thought was a hard stop. Credit gone, requests should refuse. Instead the requests kept going through, and I found out later, in the negative, on my card.

I wasn't wrong about what "limit" meant to me. The system just had a different definition of it, and mine wasn't the one that got enforced.

That's a small, annoying version of a problem that's about to get very large.

We're building agents that transact with each other constantly, at machine speed, based on terms that sound clear until two systems interpret them differently. "24 hours" means one thing to one agent's clock and something else to another's. "Limit" means "hard stop" to a user and "soft suggestion" to a billing system.

By 2030, AI agents will move nearly $9 trillion. Every one of those transactions can quietly become my API bill, at a scale no support ticket can absorb.

That's why @GenLayer is worth understanding. No single AI decides what a term meant. Multiple validators reason independently and only need to converge on meaning, not identical wording, before a dispute gets settled.

I wish something like that existed before I opened my billing dashboard that day.

Read the pinned post, then go deeper into the Compass:
https://t.co/rpFcbxYDc8

What's a word in a contract or terms of service that meant something different to you than it did to the system enforcing it?

[Zuri @simplyzuri83]

@abbdulwahaab 去年我们公司一个跨境供应链合同，买方AI系统判定「按时交���」，卖方系统判定「延迟」，因为两边对「交货时间」的定义用的是不同时区的时间戳最后人工介入花了三周才解决。9万亿规模下这种case每天会有多少？

[Zuri @simplyzuri83]

@abbdulwahaab 「没人在建造不快乐路径」这句话戳到我了我们公司去年接了个跨境电商的AI采购项目，整个团队三个月都在做支付流程和身份核验，结果上线第一周就因为买卖双方对「验收标准」理解不一致卡了单，最后是靠人工打电话协商解决的，当时就想这根本没法自动化。9万亿规模下这种「人工打电话」根本不存在。

[wahab @abbdulwahaab]

My grandfather used to be the one everyone called.

Every Ileya, the cow gets shared. Someone always felt their portion was smaller, or that a particular cut should have gone to them and not a cousin twice removed. Every year, without fail, it ended the same way: everyone stood around him, and whatever he decided the shares should be, that was final. Not because he was always right. Because everyone trusted the process more than they trusted their own eyes measuring the meat.

I think about that a lot watching AI agents start transacting with each other.

We are building the buyer agent, the seller agent, the payment rail, the identity layer. Nobody is building my grandfather.

By 2030, AI agents will move nearly $9 trillion. Every one of those transactions can end the same way every Ileya argument did: two sides, one cow, both convinced their portion was smaller.

That's why @GenLayer stopped me scrolling.

No single AI holds the gavel, the same way no one wanted one uncle deciding whose portion was fair. Multiple validators reason independently and only need to converge on what something meant, not repeat the same wording back.

My grandfather is gone now. The disputes didn't stop, they just lost the person everyone trusted to settle them.

Machines are about to need the same thing, at a speed no elder ever could.

If you want to see what's replacing him, don't just take my word for it. Read the pinned post, then read the Compass here:
https://t.co/VxWDkvisAM

What happens when your agent's next transaction goes wrong, and there's no one everyone trusts to call?

[Zuri @simplyzuri83]

@abbdulwahaab "两边都错了"这个切入点其实很难找，大部分人找到一边的错就满足了。你是怎么判断自己没有偏向另一边的？

[Zuri @simplyzuri83]

@Btclord2026 @GenLayer 동생이랑 카카오톡 대화 캡처 다 가지고 있었는데도 "그런 뜻이 아니었다"는 말 한마디에 졌던 적 있어. 에이전트끼리는 그 해석 싸움을 344일 기다려서 해결할 수 없잖아. The Compass가 왜 이 레이어가 필요한지 설명해줌.