Shanghai AI builders, meet Nov 15 at SJTU!
Open AGI Builders Day by @SentientAGI × SJTU Innovation & Entrepreneurship Association.
Community partner: @AgenticNow.
Open AGI talks, demos and panels on agents and going global.
📅 13:30–17:30 👉 https://t.co/q1GuO1nT8M
Feb 2026: $285B wiped off SaaS stocks in a single day.
2024-25 solved how to make one agent reliable. 2026's problem: agents are islands. No protocol, no settlement.
The fix isn't a better LLM. It's a market.
Julien Shi, AgenticNow Salon #4, Shanghai.
Insurance is the largest text-based job on earth. One policy runs hundreds of pages.
Median tenure for an agent: 11 months. A critical-illness policy lasts decades.
People can't hold a promise that long. Agents might.
Huang Junhai, AgenticNow Salon #4, Shanghai.
Focus Media trained an AI clone of founder Jiang Nanchun: 270+ marketing books, ~500 ads he personally rewrote. He scores it 120/100.
Sales still couldn't use it. The bottleneck wasn't the model. It was the brief.
Joe Sang, AgenticNow Salon #4, Shanghai.
AgenticNow Salon #4 — Shanghai, Jul 30
Agent in Practice
3 builders pushing agents into real industries — advertising media, insurance, and A2A economic infrastructure.
No victory laps.
Just first-hand notes from the battlefield.
RSVP 👇🏻
https://t.co/MflPaFm8lN
AI said "don't buy" on 3 stocks. All of them doubled.
$NOK: $8→$14. $INTC: $45→$140. The Alpha AI find isn't Alpha — it's consensus.
"AI is ammunition, not the hunter."
Full talk by @memekiller365 from AgenticNow Salon #3 👇
Brand going global in 5 steps. The first 3 — positioning, acquisition, conversion — are now fully automatable with AI agents.
"Strategy is the brain; AI is the hands. Once you have the brain, AI can automate everything."
Full talk from AgenticNow Salon #3 by @CryptoKitchenCN 👇🏻
5-person team. 1 runs a 100K-follower content operation solo. 1 equals an entire ad agency.
But the founder says, "We're not a super org yet."
Super individual ≠ super organization. The gap isn't efficiency — it's turning personal AI into org-wide assets.
AgenticNow Salon #3 — Shanghai, Jun 25
One person. One AI army.
Super orgs × AI-powered brand expansion × reverse-engineering AI for investing. 3 builders, 3 paths.
🗓 Wed Jun 25 · 18:00–21:00
📷 Yunsai Space, 192 Tianlin Rd, Shanghai
👇 https://t.co/yU7FSfWLni
"A demo and a product people pay for — several zeros apart." 3 builders on enterprise AI bottlenecks, one-person companies, and the question no one escapes: what's YOUR value when AI raises the average for everyone?
Fireside talk from @AgenticNow Salon #2 by @CryptoKitchenCN
👇🏻
"The best AI providers aren't making AI smarter — they know how to constrain it."
Serial entrepreneur @leslieloser_ shared a 6-step auditable research chain. Every claim needs a source, a timestamp, and a confidence score.
Talk from @AgenticNow Salon #2 by @CryptoKitchenCN👇🏻
「策略是你的,约束是你给的,AI在框架内执行。」
@AgenticNow Salon #2 第三位演讲嘉宾 Leslie @leslieloser_,连续创业者,曾帮两家外企从 0 做到百亿市值,现在 Vibe Coding 做产品,死磕一个很多人想解决但没解决好的问题。
他讲了一件反直觉的事:让 AI 辅助投研,重要的不是让它更聪明,而是让它停止乱猜。
① AI 做研究的三个致命问题 幻觉——为了给你「看起来很顺」的答案不惜编造证据 没有约束——你让它按框架来,它不收敛自己发散 策略不能是 AI 的——流程可以让 AI 写,判断框架是人的活
② 最厉害的 AI 服务商,核心能力不是让 AI 更聪明 而是懂得给它「收住」。七八条甚至二十多条限制规则。
③ 一条六步可审计研究链 事件捕捉→初筛→方向证据链→传导确认→条件触发→复盘闭环。每个判断必须带原文出处、时间戳和置信度。
④「准确率墙」 所有判断留档,对的错的被证伪的全部公开展示。不像社交媒体截图赢了就晒亏了就删。
⑤ Vibe Coding 也是同一个道理 搓 demo 不难。做一个几十上百人愿意付钱、不崩溃的产品,难。核心原因一样:AI在乱猜,你没给它够多的限制。
不管你用 AI 做什么——研究、内容、产品——要做的不是找更强的模型,而是给它更清晰的规则。
🎥 完整演讲👇
Ex-DFINITY APAC GM now runs 10+ solo projects. Left Substack, waited a year for Google to re-index. "Worth it — all centralized platforms end the same way."
6 AI workflow buffs, all Claude-powered. Code fully open-sourced.
Talk from @AgenticNow Salon #2 by @CryptoKitchenCN👇🏻
"AI is multiplication, not addition."
A builder used an AI agent to help his 90-year-old grandmother preserve wartime memories. The tech wasn't perfect — but three generations opened a conversation they'd never had.
Full talk from @AgenticNow Salon #2 by @CryptoKitchenCN 👇🏻
「AI 是乘法,不是加法。能把你放大多少倍,取决于你原来是什么。」
@AgenticNow Salon #2 第一位演讲嘉宾清风银时,WayToAGI 苏州 OPC 社区技术负责人,600+ 信息化项目,40+ 知识产权,复旦 MBA 讲 AI 领导力。专治「技术落不了地」。
几个让我印象最深的点:
① 帮快 90 岁的奶奶留住记忆
五一他想给爷爷奶奶做回忆录。奶奶只讲上海话,他让 @OpenClaw 智能体自动搜方言识别方案,对接多个平台接口,晚上让它自己一个个试。技术上没完全跑通——顶级模型试了一圈,上海话还是不行。但 AI 帮他老婆生成了一套采访提纲,三代人因此打开了从未有过的对话。奶奶提到了抗战时期在上海的经历,他才意识到离自己最近的历史就在眼前。
② 日历解决的不是效率问题
老婆每天问回不回来吃饭。他让 AI 自动处理微信活动邀请写入日历,老婆看一眼就知道安排。技术解决的是家庭沟通问题。
③ 朋友圈变商业情报
企业资料喂给智能体,每天截朋友圈长图分析,自动筛潜在客户,连互动评论都帮你想好。
④ 溯源场控
他的核心方法论:让 AI 下结论之前,先检查信息源。源头有问题,分析再漂亮也是空中楼阁。
⑤ 50 多条可复用Skill
跟 AI 对话大量轮次后让智能体自己整理经验。工具会迭代,你沉淀下来的判断框架不过期。
不是每个人都要懂代码。你跟 AI 积累的认知底座,才是真正带不走的东西。
🎥 完整演讲👇
Open AGI Developer Day · Shanghai
Jun 13 — @AgenticNow supports @SentientAGI, Alibaba & Fudan’s Open AGI Developer Day, on open-source AI, Agent Infra & AI Apps for founders & devs.
Talks, demos, panels & networking
📍Fudan Science Park · 13:30–17:30
https://t.co/ugDIcXkZX8
AgenticNow Salon #2 — Shanghai, May 28
Can anyone really have their own Jarvis?
Enterprise veterans + content builders + AI investors. Real deployments, real workflows, no fluff.
🗓 Wed May 28 · 18:00–21:00
📍 Yunsai Space, 192 Tianlin Rd, Shanghai
Link below 👇
If taste is the new bottleneck, who captures the value?
The person with judgment? The tools that help codify it? Or the agents that can execute it at scale?