Yang Song

my dream home definitely includes a tiny coffee corner and slow little morning rituals ☕️

@chris_j_paxton Agree. Software, especially middle ware, has never been the focus in robotics field. Compared with algorithms, sensors, and models

Again robotics software sucks. Its so bad. Part of what makes it so bad -- and what makes stuft like ROS useful -- is the amount of boring middleware code you need to write between sensors, processes, etc. Coding models make it so much easier to work with, its wonderful.

All the venture money in robotics is going to three verticals: 1/ The robot body (Figure, 1X) 2/ The robot brain (Physical Intelligence, Skild AI) 3/ The infrastructure around the brain, mainly data collection, simulation, pre-training stacks If you're now starting a company in any of these three, one of two things has to be true: > You have a PhD from a top-institution in robotics and did world-class research over the last ten years. > You have a sharper insight into what would work, and articulate it very clearly. If none of these are true, you are probably better off spending your time on deployment. Essentially, become the operator who knows what the robot has to do, in what order, against which constraints to earn its position within a vertical. If you can do that, I think there's two clear outcomes: 1/ You build the robot, because your customer insight is so specific that you can only build that robot. 2/ The frontier labs acquire you for the relationship. They have the model. They need the customer. You have built the customer relationship over the years.

@BijanTavassoli @aru9a3 Quote from my Japanese friend: “that is my place”. I can relate, like Chinatown in NYC

@ysongtwitt @aru9a3 Never been there and I‘m German but heard good things😅

ドイツに無事着いたのですが、アジア人が少なくて（当たり前）、午後6時でも昼間のように明るくて、日差しが強いけど湿気ってなくて、売ってるものが全然ちがくて、言葉も全然わからなくて、でもつたない英語で道を聞いたらみんな優しく教えてくれて「Have a nice day!」って言ってくれてすごく嬉しくて、なんだか自分が自分ではないみたいなフワフワした気分で、そして何となく買ったサラダが山盛りで、豆が超いっぱい入っていました

@MarioNawfal right. The best way is not solving a problem but stop creating the problem

🇨🇳 China has developed autonomous robots to clean pollution from rivers and lakes Ingenious, but maybe stop polluting them as well, eh?

In a recent batch talk, YC General Partner @t_blom broke down how to build a self-improving, AI-native company. He walks through how to create recursive, self-improving AI loops, and why founders who get this right will run companies that improve while they sleep. 00:00 — Companies Are Roman Legions 00:54 — Copilots Are the Wrong Mental Model 01:55 — Extract the Domain Knowledge 02:24 — The Recursive Self-Improving Loop 04:12 — The Holy Shit Moment at YC 05:50 — Self-Optimizing Product and Support Loops 06:29 — Burn Tokens, Not Headcount 07:23 — Middle Management Is Over 08:05 — Make Everything Legible to AI 09:40 — Regenerating the YC User Manual 11:19 — Software Is Ephemeral, Context Is Valuable 12:18 — Where Humans Still Matter

@Lady_Iona @ChinaEV_Eng_Lif Not really. Actually what I experienced were mostly gas cars parked there. No penalties

@ysongtwitt @ChinaEV_Eng_Lif Here in Germany you must pay a hefty blocking fee, if you do no leave the charger when charging is finished. But, but to trash a car: No way!

It’s wrong what he’s doing. However, it’s relatable.

COOL ! 有人拿 新发布Qwen 3.7-Max和Claude Opus 4.7、GPT-5.5 做 Agent 循环对比：让模型自己写俄罗斯方块机器人、自己测试、连续迭代 10 轮后直接 PK，结果： Qwen 3.7-Max：+$56%，成本 $1.32 Opus 4.7：+$28%，成本 $12.15 GPT-5.5：+$7%，成本 $2.85 x.com/atomic_chat_hq…

@ScottyIsAbroad @ChinaEV_Eng_Lif Different. Once man ever drives an EV, there is no come back

@ysongtwitt @ChinaEV_Eng_Lif Get an ICE car. Problem solved.

One more thing 🚀 Qwen’s agentic capability is no longer limited to the digital world — we’re bringing it into physical world. With our in-house robotic agentic system and navigation model, Qwen can now control a robot to execute tasks in real-world. #qwen #embodied #robotics

Machine learning…….

@GermanSimply_ ahha, could not agree more. the automated phone menu always KOs me. It is hard to dial the phone meanwhile using google translate app, so I created my local app for this using LLM.

𝟳. Your German will be tested daily – in ways no textbook prepared you for. The bakery. The doctor. The Bavarian neighbour. The automated phone menu at impossible speed. 𝘏𝘰𝘤𝘩𝘥𝘦𝘶𝘵𝘴𝘤𝘩 gets you through. Exposure is the only real solution. Start before you arrive.

Everyone tells you about Germany's efficiency. The punctuality. The engineering. The order. Nobody warns you about the other things. The things you only discover after you've already moved there. 10 of them. Right here. 🧵🇩🇪

@sutoroveli_news Only certain groups of customers like soldiers and monks can be served right?

【誕生】客の頭をスキャンして自動で髪を切るロボットが街に登場した。3Dスキャンで顔の形と髪質と希望スタイルを分析しロボットアームがミリ単位で正確に切る。中国の都市部で1回約60円という低価格で複数の理髪キオスクが運用されている。

@tokufxug bravo. Once people in the AD field realize its true power, 80% of repetitive data infra pains could be cured

自動運転データセットを扱うオープンソースライブラリ「123D」。 KITTIやWaymo等の異なるデータ形式を統一し、センサーやHDマップの共通表現を提供。 Apache Arrow採用で高速動作し、3,300時間の走行記録で評価や共同訓練を効率化します。（リンクはリプ欄）

Once it becomes a new data layer standard, it boosts the embodied ai development. 模型决定上限，数据基础设施决定能否真正到达那个上限。

1/6 Introducing Qwen3.5-LiveTranslate: Next-gen real-time interpretation is here. 🌍 We’re breaking down language barriers with 3,500+ language pairs, ultra-low latency, visual context, real-time voice cloning, and hotword customization. Engineered to help you ship native, frictionless real-time translation experiences to a global audience.

holy shit😭 这绝逼是2026年顶级产品该有的营销,3秒胜过30页白皮书！ 一个卖数据层的公司， 用一个机械臂彩蛋， 把所有B2B营销按在地上摩擦， 把最硬核的概念变成最上瘾的玩具,真的太炸了💥 以后所有B2B公司的官网 都应该被这个机械臂枪毙一遍！

Mom is sleeping; she's tired. 🤫 🥰