Simon Shg @ShengSSX

Building @parall_hq parall.com SF Joined March 2023

Tweets

65
Followers

55
Following

67
Likes

135

Simon Shg @ShengSSX

2 days ago

@ok1mraise @Parall_HQ @typesafeai but yea the intent filtering is the key, I'm using it to pass my hundreds of emails/day to different team agent. works great

1 0 1 26 0

View Details

Simon Shg @ShengSSX

2 days ago

We at @Parall_HQ are open-sourcing a universal agentic monitor built around Jev (@typesafeai). It’s a fun little toy, but a lot handier than expected. ➡️ github.com/parall-hq/jeva…

2 1 9 706 4

View Details

you'd want both, don't you 😂 a rule can mix hard filters with Jev judgments, e.g. line.contains("outbound") && judge.boolean(line, "Is this person describing a problem we solve?") >= 0.7 hard checks run first, so Jev only sees what's left. jevable test finds the threshold from a few yes/no examples

0 0 1 27 0

View Details

Simon Shg @ShengSSX

2 days ago

give your agent Jev with one line of jevable, our open source CLI. paste this into your agent: Set up jevable: run `npx -y jevable guide` and follow it ↓ github.com/parall-hq/jeva…

0 0 1 61 1

View Details

Simon Shg @ShengSSX

2 days ago

2 AI colleagues just booked us a demo off someone's random post online Ava and Owen are the two agents running our prospecting in @Parall_HQ. they live in our @Parall_HQ workspace like everyone else on the sales team, and @typesafeai Jev makes the quick judgement calls for them > Jev reads every post, Ava wakes up only for the ones that fit us > Jev rates every account, Ava only works the best ones > Owen writes the email, our human AE approves > Jev sorts the replies, Owen only chases the interested ones you don’t need to hand them leads. jevable makes it possible for proactive agents to own a job at 86% less cost than agents that check in every 30 min

Simon Shg @ShengSSX

2 days ago

We at @Parall_HQ are open-sourcing a universal agentic monitor built around Jev (@typesafeai). It’s a fun little toy, but a lot handier than expected. ➡️ github.com/parall-hq/jeva…

2 1 9 706 4

1 0 3 143 1

View Details

Simon Shg @ShengSSX

2 weeks ago

@yan5xu @indigox 你哪儿有“被迫快速交付”😤

1 0 0 20 0

View Details

Simon Shg @ShengSSX

3 weeks ago

Jev is the most mind-blowing thing happened this year, and it will change far more than anything else has. Can't imagine how many "can't-dos" are now possible. Insanely fast, insanely cheap.

0 0 0 33 0

View Details

Simon Shg @ShengSSX

3 weeks ago

@notkevinzhang Use as a verifier for reward signals in human-ai coordination harness

0 0 0 10 0

View Details

Simon Shg @ShengSSX

2 months ago

Last night, I joined Yu Yi and Wang Wei at @geekpark to discuss the rise of Forward Deployed Engineers. Here’s my take: FDEs are booming not because models are still too weak, but because they’re finally strong enough. (Stronger than enterprise can imagine) We used to think AI couldn’t make its way into enterprises because the intelligence wasn’t there yet. Now it’s becoming clear that the bottleneck has moved—from “can the model do the work?” to “does the organization know how to let it?” Giving a company a model or an API doesn’t automatically create value. How roles are designed, how humans and agents divide the work, how permissions are granted, how workflows are rebuilt, and how learnings are retained—these are what determine whether AI can truly become part of the business. So to me, FDE is neither on-site outsourcing nor simply helping customers install a product. It is a process of discovering the best organizational practices of the AI era together with customers, then encoding those practices back into the product and the agent team. The best FDE practice shouldn’t grow linearly with every new customer. After each deployment, the next one should require fewer people—until the agent system can handle most of the deployment and evolution itself. Model companies are creating intelligence. We want to turn intelligence into an organization. In a sense, this is what we mean by accelerating AGI’s transition into organizational intelligence.

1 0 3 3K 1

View Details

yan5xu @yan5xu

5 months ago

两个月前说要做大学生 AI 资助项目，现在要落地啦。解字计划 — 取自《说文解字》。字是语言最小单位，token 是大模型最小单位。项目资助大学生获得 AI token。第一年我个人出资 13 万+，预计资助 600 名大学生。计划草案已公开，欢迎 review： github.com/jiezi-ai/grant 特别想听听： 1. 阶段设计合不合理 2. 筛选门槛会不会太高/太低 3. 有没有遗漏的关键问题预计5.15 定稿，希望大家能多提提意见

yan5xu @yan5xu

6 months ago

Pin 一下。春节时就在规划，希望能够做一个资助项目，给大学生100～1000的token，帮助他们去接触AI。两周前还和一个大哥说到这个事情。希望能在四月底之前把这个资助项目跑起来。

12 3 44 59K 10

20 4 40 18K 20

View Details

Simon Shg @ShengSSX

5 months ago

@GergelyOrosz Coinbase had an outage right when everyone wanted to trade 😂 timing could not be worse

0 0 0 129 0

View Details

Simon Shg @ShengSSX

5 months ago

Builders don't have a guild. That's the point.Doctors and lawyers will be protected by policy. Coders and founders won't — which means we're the first real test of what happens when AI removes the legal floor. Scarier? Maybe. But also: no one's forcing us to fight yesterday's battle.

0 0 1 195 1

View Details

Simon Shg @ShengSSX

5 months ago

The next wave of AI products won't be won on model quality alone. They'll be won on response time, conversational feel, and how naturally they fit into human workflows. The bottleneck is shifting from "what can AI do" to "how does interacting with AI feel.

0 0 0 16 0

View Details

Simon Shg @ShengSSX

5 months ago

This is the part builders miss: the "last mile" of AI UX is often physical infrastructure. You can have the smartest model in the world. If the pipe is slow, the experience is bad. Human-AI collaboration breaks down at the interaction layer, not the intelligence layer.

1 0 0 20 0

View Details

Simon Shg @ShengSSX

5 months ago

@yan5xu 对而且直接自己就带着plan来问我干不干了

0 0 0 39 0

View Details

Simon Shg @ShengSSX

6 months ago

看板的本质不是那个GUI，是组织信息流动的结构。这个需求不会消失。多人+多agent并行工作时，任务之间的关联、依赖、状态更需要一个中心化的地方来对齐，否则context分布式散落在每个agent本地，协同就是断的。（也不安全）我们认为看板要分化成两层：人侧还是可视化的board（人需要直觉友好的界面），AI侧是一个中心化的context流转层——task、知识、讨论在同一套数据模型里原生打通，agent直接从结构化context工作，不用从碎片消息流里猜。