What's actually happening in China AI from the inside. 600K+ builder community. Substack https://teapawt.org/t.co/b2Ml0b8gHK
Joined January 2026
- Tweets216
- Following371
- Followers93
- Likes72
Pinned Tweet
We're excited to announce the launch of… a 16-year-old media company.
Today @UnitreeRobotics rang the bell on the STAR Market. In 2018, when the entire robotics sector had given up, our fund wired them ¥2M with no signed contract — the first institutional money Unitree ever took. Wang Xingxing was paying salaries out of his own savings at the time and hadn't told anyone.
Founder Park is the AI arm of GeekPark. Since 2010 our whole job has been finding founders a few years before the market agrees they matter:
→ Lei Jun, when Xiaomi was a phone company with a forum
→ Zhang Yiming, when ByteDance was a news app
→ Wang Xingxing, 2017, with a robot dog nobody had a category name for
→ Xiao Hong, 2023, when he was doing Monica before Manus
So no, we're not a new company. We're just new in English.
v1.0 ships today. Sixteen years late, in a language we don't publish in yet. Classic us.
Founder Park retweeted
@Recursive_SI shipped its first model, Flint! Exciting to see they did extensive RL training on the Pi harness, so it works well out of the box.
40B MoE (2.5B active) for fast local coding. Apache 2.0.
Available now for free in Pi via Radius.
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
Founder Park retweeted
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
Founder Park retweeted
花费16天，我用 Opus 5.5 将小王子的故事制作成了3D剧情游戏
结合Tripo+Blender，Opus 5.5 将小王子故事中的每一个星球变成了可探索的3D场景
希望大家能支持一手，后续会开源整个游戏，以及详细制作流程
Welcome to The Little Prince: Across the Stars✨：littleprince.dev/
Founder Park retweeted
$52M seed round. 15M+ users. But this is what we're most proud of yet.
Introducing Drama 3: the most emotive and controllable TTS model in the world
And yes, it's open-weight.
You can now control how even a single word should sound, change its tone, and shift your voice mid-sentence.
Uncle Roger was impressed too.
Try it here:
s.fish.audio/rqh0l8
Comment “DRAMA” within the next 48 hours for 1 month free.
If you're an enterprise your first month is on us:
s.fish.audio/qz9r0t
You can watch a pitch online. You can’t watch a founder think on their feet through a screen.
Founder Show is in person this time. Singapore, Oct 30. 8 founders, 10 minutes each, then open questions from the room.
Come ask.
Founder Park retweeted
Introducing Hypit: Clone any viral video with AI agents.
1 clone, 100 variants, 100M views.
Hypit lets your AI agent (Claude Code, Codex...) clone any viral video.
Paste any viral video link from TikTok, Instagram, or YouTube into your agent. Hypit clones it into a complete agentic video workflow: footage, captions, B-roll, effects.
GitHub: github.com/hypit-ai/hypit
Key Point:
- Arcads: $220 / month
- Higgsfield: $129 / month
- Creatify: $99 / month
- Hypit: FREE 🌟
Build the video creation harness for AI Agents.
Redefine how videos are made.
Like + comment "Hypit" and I' ll send you 100 free AI avatars, each with its own unique voice.
And if you like what we’re building, give Hypit a 🌟 on GitHub.
Astra. Fable 5.1. Big week for the model layer.
Sep 17 we're running a pitch session for the layer above it.
Apply to pitch or register to watch 👇
9/12 周六下午，北京，线下闭门 Idea Rush。
规则只有一条：全场不许说「但是」。
5-6 个人各带一个 idea，8 分钟讲完。全场轮流开麦，只能干三件事：帮 TA 加功能、帮 TA 找场景、帮 TA 想怎么实现。
带 idea 来或者带添砖加瓦的心态来都行
报名 👇
Founder Park retweeted
The full EVA01 paper is now live.
For a long time, Text-to-3D has largely followed one path: Text → Image → 3D reconstruction.
It can produce impressive 3D assets, but there is a fundamental gap. The model that understands the prompt does not directly reason about the final geometry. Semantic understanding and geometric generation happen in separate stages.
For one-shot asset generation, that can be enough. For games, it quickly becomes a limitation.
EVA01 was built around this idea:
From AI that generates 3D
to AI that can understand and work in 3D.
Read the full EVA01 paper on arXiv:
arxiv.org/abs/2605.16745
Founder Park retweeted
Since announcing Jalapeño, our first custom inference chip, we’ve been testing it and the system around it.
The results show a major advance: more intelligence from every watt and faster responses, delivering both higher throughput and lower latency in one architecture without sacrificing efficiency.
A shared, cross-agent standard for config files and skills is table stakes
Founder Park retweeted
If you're curious about what's happening with the new harness in Pi, a lot of activity is taking place in the dev branch and not on main :)
