Jeremy Wen

@CarnegieMellon @teppercmu
previously
- PM @ TikTok
- VC @ top social app (investment thesis failed)

[░ perfectloop ░ @PERFECTL00P]

𝘿𝙀𝘼𝙇 𝙒𝙄𝙏𝙃 𝙄𝙏 🕶️

[AlloomiAI @AlloomiAI]

1/ Most AI agents do not fail because they lack actions.

They fail because they forget the work.

In practice, that shows up as four memory problems: no structure, no sense of time, too much noise, and no feedback loop.

OpenLoomi is built to fix those four gaps.

[Jeremy Wen @Jeremy__Wen]

we are building 8 out of 10 agents mentioned in the post, if you want to have a dedicated tech team to help you automate your workflow, talk to me or download our app at
https://t.co/BBia43lNsQ

[Shann³ @shannholmberg]

build your revenue-generating ai systems first

here are 10 agents I would build before any content or ops agent, each one wired into the same company brain and CRM:

> instant sales call feedback. the rep hangs up and an agent has already pulled the transcript, scored the call against your framework, and sent back what they missed, which objection they fumbled, and the line to use next time. reps get better between calls instead of in a quarterly review

> proposal generator. pulls the call transcript, scrapes the prospects site and context, layers in your sales managers notes and the closest past wins, and drafts the proposal in your format. the rep edits instead of starting from a blank page, and it goes out same day instead of three days later

> inbound lead qualification. a lead comes in, the agent enriches it, scores fit, and replies in minutes. if they qualify it either gives them real value straight away or drops a direct link to book a meeting, so you never lose a hot lead to a slow inbox

> CRM agent. has company-wide context and keeps the pipeline clean, enriches every record, surfaces insights, and nudges reps when a deal goes stale or a step gets skipped. it keeps the whole sales team consistent instead of each rep running their own system in their head

> automated follow-ups. every lead that goes quiet gets a full sequence, not one "just checking in." the agent writes each touch off the actual context of the deal, so the follow-up adds something instead of nagging

> contract generator. takes the agreed terms off the deal, drafts the contract in your template, flags anything off-standard, and routes it for signature. cuts the gap between verbal yes and signed from days to minutes

> upsell agent. watches your company brain and CRM for the moment a client is ready for more, hit a usage cap, mentioned a new goal on a call, growing fast, and hands the account owner the opening with the context to act on it

> churn fire-fighting agent. a head of accounts that spots a client going cold before they tell you, slow replies, dropped meetings, a short reply that used to be warm, and loops in the team to save it while theres still time, not after the cancel email lands

> lead reactivation agent. goes back through lost and ghosted deals, enriches each one with whats changed since, builds a plan to reopen it, then executes. your dead pipeline becomes a list the agent works on its own

> invoice payment follow-ups. an agent owns the dates, knows when each client should pay, and starts the follow-up the moment its late, so you stop financing clients for free and cash lands faster with nobody chasing it by hand

[Tesla Owners Silicon Valley @teslaownersSV]

Engineering is magic

[jietang @jietang]

Recent thoughts:
The Shift to Long-Horizon Tasks
The most likely breakthrough this year will be in long-horizon tasks. We are moving toward a stage where Large Language Models (LLMs) learn to complete extended, complex missions by interacting with Agent environments. This is perhaps where the true value of LLMs lies. Take cybersecurity as an example: imagine a model that continuously hunts for software bugs and vulnerabilities. While it sounds like a search process, it’s actually the model learning the high-level intuition and methodology of a professional hacker. Unlike humans, AI can run 24/7 without fatigue. It could potentially find exploits at a much higher frequwill ency and claim bounties on platforms like HackerOne or BugCrowd. It sounds fun, but fundamentally, it's a revolution that displaces the hacker. If even hackers are being "disrupted," one can only imagine the impact on general programmers.

From One-Person to None-Person Companies
Building on long-horizon capabilities, Autonomous Agent Systems (AAS) will inevitably become the next frontier. Last year, we were discussing the rise of the "One Person Company" (OPC). I didn't expect us to move so quickly toward the "None Person Company" (NPC). It’s an ironic twist—we might all end up as NPCs in this new ecosystem.

Engineering the Impossible: Memory and Learning
To realize the vision above, we must solve three technical pillars: Memory, Continual Learning, and Self-Judging.
I used to think these would require massive paradigm shifts and years of research. However, the pressure from both the technical and application sides is so intense that we are seeing these capabilities emerge through ingenious engineering "tricks":
Memory: Long context windows (1M+) and RAG have significantly bridged the gap.
Continual Learning: While true continual learning remains difficult, the release cycles are shrinking. Global models are updated monthly; domestic models are catching up. If we reach weekly updates by next year, it will effectively function as continual learning.
Self-Judging: This remains the most elusive, yet models like Opus 4.7 are already demonstrating early self-correction and judgment capabilities.

The Self-Evolving Endgame
The most difficult—and most promising—path is Self-Evolution. The current wave is incredibly fierce. I suspect that models like Claude may have already achieved a baseline for self-training: writing their own code, cleaning their own data, generating synthetic data, and then training on it. It might "waste" some compute, but it saves the most precious resources: human labor and time. In the LLM era, speed is everything. Rapid iteration is what creates the cognitive gap between leaders and followers. Claude’s rumored 2-million-chip cluster for next year is likely dedicated to exactly this: autonomous model self-training.
Technical Summary:
1M Context: Necessary baseline.
Memory & Continual Learning: Prerequisites, likely solved first via "tricky" engineering.
Harnessing Environments: The breakthrough point.
Self-Judging: The tipping point.
Full Self-Training: The endgame.

Redefining AGI and the Industry
If this is the road to AGI, then AGI’s definition should be the sum of all human collective intelligence, not just an individual’s intelligence. It must possess the creative capacity to produce something as profound as the "Theory of Relativity"—meeting the bar set by Hassabis.
During this transition, every APP will need to be reconstructed as AI-native. In fact, we might move past the concept of APPs entirely. The most significant challenge will be the reconstruction of the operating system itself. In the future, you won’t see a traditional desktop; you will see an LLM OS, where applications are "generated on demand." This challenges the 80-year-old Von Neumann architecture and represents a total upheaval of the computer science industry.

The Irreversible Wave
From completing long-horizon tasks to fully autonomous operations, every sector—Security, Finance, Law, E-commerce—will be reshaped. Many friends have reached out lately, asking how to transform their enterprises to keep pace with AI. But few truly realize that this irreversible process has already begun. As this massive technical wave hits, we must be prepared to act, but we must also start thinking seriously about how to regulate it.

[Lindy @getlindy]

https://t.co/yd56zTPayy

[Jeremy Wen @Jeremy__Wen]

@kathrynwu1 @openmartai great product Kath!!

[Kathryn Wu @kathrynwu1]

Today we launched OpenClaw for Sales.

It combines @openmartai’s data, LinkedIn data, and top data sources to do the work for you.

Used by teams like Whatnot, DoorDash, Alibaba, and many others.

Try for free: https://t.co/WdJe81mD8j or DM me for an invite code.

[Teddy Riker @teddy_riker]

https://t.co/dyiVwtFQNJ

[Jeremy Wen @Jeremy__Wen]

clicky is such a bad ass

[Farza 🇵🇰🇺🇸 @FarzaTV]

Here's the new Clicky.

It's the simplest interface in the world to talk to AI + spawn agents.

It builds Mac apps. It does research to help you find IG micro-influencers. It interacts with native Apple Notes, Calendar, Reminders.

Built for consumers, 0 setup.

Try today, free.

[Jeremy Wen @Jeremy__Wen]

@NousResearch how did you guys make each release so aesthetically enjoyable

[Silicon Mania @siliconmania]

last week in tech was based.

[Zain Shah @zan2434]

Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see.

@eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)

[Nous Research @NousResearch]

By popular demand, here's the full version of the song - along with a music video created entirely by Hermes Agent.

Enjoy "The Ballad of Hermes":

[Shiv @shivsakhuja]

Introducing Gooseworks — AI coworkers that can do real GTM work.

Over the last few months, we've been obsessed with turning AI into actual coworkers for our team.

And it's working.

Goose is now our most valuable teammate – it finds high-intent leads, runs outbound campaigns, coordinates with influencers, tracks SEO / AEO and a ton more.

Today we're launching this publicly so everyone can create their own AI coworkers.

Each coworker (or Goose) has its own computer, filesystem, mailbox, memory, and tools.

Goose can work autonomously for hours, communicates over Slack, iMessage, WhatsApp or email, and gets smarter as you give it more context.

It's not just a personal assistant — your whole team can talk to Goose from anywhere, and schedule it to run on autopilot.

And here's the best part – we've already given Goose 100+ skills and data APIs for GTM work:
– map your TAM
– search people databases with natural language
– find leads from X and LinkedIn posts
– monitor intent signals from your ICPs
– update your CRM and generate reports
– find influencers and run outreach campaigns
– track visibility in search and answer engines
– connect to all the SaaS tools your business runs on

+++ lots more.

We also open-sourced every skill as a toolkit you can install directly in Claude Code.

Comment "Goose" and follow and I'll send you the install link.

[Jeremy Wen @Jeremy__Wen]

@Watch_Velocity its art

[醉清風 @Xin78787]

日文和草书对比，他们回去的路上让海水给字泡没了……

[Jeremy Wen @Jeremy__Wen]

如果纯粹从投资的角度出发，那么是不是应该关注那些一些在小红书生态（中国）和X生态（海外）做出了小工具解决了某个细分场景的痛点然后帖子获得大量转发/评论的原生创作/创业者？

我先来推荐一个 @FarzaTV

[huangyihe @huangyihe]

1、中国创始人更像“已知问题的顶级执行者”，而不是提出没人想过的问题的人。历史上最成功的中国创始人（如马云、任正非、刘强东）反而是“非典型”路线，现在的体系却在筛选掉这类人。

2、深圳的硬件“地下工厂”令人震撼。工程师把西方高端产品拆解、逆向工程，供应链高度集中（70%以上硬件输入来自大湾区）。迭代速度远超西方。很多公司走DJI路线——先做小众消费硬件（电动轮椅、割草机器人等），做到数亿美金营收，再横向扩张。

3、开源模型很厉害，但闭源大模型仍大幅落后西方，差距可能扩大（GPU受限、蒸馏被西方封堵、CapEx差距巨大）。

4、软件创业多是前字节PM/研究员，抄西方agentic/ambient消费软件，针对海外市场，但很容易被大模型实验室一键取代。

5、几乎所有创始人都在用Claude、关注Dwarkesh播客，对硅谷生态了如指掌，且优先做全球市场（尤其是西方），而非先做中国市场。

6、真正值得投资的是那些不走大厂学历路线的“异类”创始人，目前本地VC很少在挖这类人。

[𝐬𝐨𝐯 ᯓ ✈︎ @sovietsoleri]

Watching The Killer (1989) on the big screen tonight, gangster-cop drama directed by John Woo. Some intoxicating visuals.

[Jeremy Wen @Jeremy__Wen]

@FarzaTV this is super helpful! where do i find this tool