Narit B

aroi

[Narit B @NaritB1]

@FreeCADNews Sure!

[Narit B @NaritB1]

In the era that we don't need to pay for high price software. Just knowing basic principles, that would be fine.

I have to implement a basic magnetic simulation plug in for Freecad, which based on MagPySim lib.

[Narit B @NaritB1]

@TJ_Bridge_AI Thx

[たろう @papa_papa_fire]

ここまで公開するのか…。

Unitree G1の54ページに及ぶ分解レポートが衝撃的でした。

中国で話題になっているUnitree G1の54ページに及ぶ分解レポートを読んでみた。

原価構成、部品、サプライヤー、設計思想、評価、弱点まで記載されています。

価格・コスト
・販売価格：約170万円
・部品原価：約83万円
・加工・組立など：約6万円
・推定粗利率：約41%

部品原価の内訳
① 関節モジュール：約66%
② センサー：約14%
③ その他（配線・コネクタなど）：約15%

産業用ロボットも関節部のコスト比率が6割前後で、ヒューマノイドもやはり同じ構成。

さらに興味深かったのは、主要部品の多くを外部サプライヤーから調達されており、ほぼオール外注で作られていました。
ハードウェア以上に、システムインテグレーションやソフトウェアが競争力になることを改めて感じました。

そして個人的に気になっていた減速機。

ハーモニックではなく、二段遊星減速機でした。分解された小型関節モジュールでは、歯数や総減速比20.58まで公開されています。思っていたより低い減速比で驚きでした。

一方で、レポートで課題として挙げられていたのは手首の可搬重量約2kg。産業用途では少し物足りない印象です。例えば、ファナックの協働ロボットCRXシリーズでも最小機種は3kg可搬です。

個人的には、手首だけ剛性や高い減速比を確保しやすいハーモニックを採用する選択肢もあるのでは？と思いました。

ただ、レポートではパワシリが本命という考え方でした。減速機ではなく、関節構造そのものが変わっていくのかもしれません。

よければ、皆さんもレポートを読んでみてください。

[Haoyu Xiong @Haoyu_Xiong_]

Success rate has long been the primary metric for evaluating robot manipulation. What about speed?

Today, we introduce ⚡️B-spline Policy (BSP). Instead of predicting discrete fixed-rate action chunks, we parameterize actions as continuous B-spline curves. Together with our system design, BSP enables fast manipulation on low-cost robot arms.

This project is co-led by @xshenhan, check out his following threads for more details. 🧵

PS: one of my favorite parts of this project was the first time we saw the robots move significantly faster and smoother than the baselines. The videos below are all real time. 👇

[jacobruh 🚀 @jrxcket]

The Chinese method of catching boosters using the wires is actually kinda cool lmao

[Eren Chen @ErenChenAI]

A Chinese creator built a 9-ton robotic arm that can be teleoperated, then used it to slap a car into retirement.

[Narit B @NaritB1]

@nexta_tv Water fighting, greeting from Thailand

[NEXTA @nexta_tv]

Berlin police turned water cannons on citizens… to cool them down

Instead of dispersing crowds, officers used water cannons to give people a refreshing shower right in the city center.

[Steven Cheng @stevencheng]

Unitree Qmini is a fully open-sourced, low-cost bipedal robot that can be entirely 3D printed by individual users.

[Narit B @NaritB1]

@JesusFerna7026 Political, and we dont think we are the owners of this country.

[空空｜预测世界杯就在Gate @btcoindown]

卧槽，黄仁勋就这么坐地上喝上啤酒了，不知道的还以为是流浪汉呢。

[Narit B @NaritB1]

@btcoindown This is what happened, when you got infinite token of AI.

[Narit B @NaritB1]

@therealcarlin Silicone has two type of curing. If you use a 1:1 or 1:2 two components type, it will be a problem. Try platinum cure. SLA usually has 5% uncured resin. Also, it long chains plasticiser. However, it the free rad that race to inhibit Silicone curing.

[pongsagon vichit @MattDIYgraphics]

Every pixel needs a depth test, but RAM is slow.

TinyGPU v3.0 hides that latency by running shader execution in parallel with Z-buffer access.

That jump—from a fixed pipeline in v2.0 to a 31-instruction pixel shader core in v3.0—happens without dropping a single frame 🚀

[Narit B @NaritB1]

@hoops18888 @malachimaxeyusa @dom_lucre the goldsmith was accused that he stolen the 25oz of gold. However, it was planted inside the epic of the building and was dissambled to proof his innocent.

[Narit B @NaritB1]

@malachimaxeyusa chang tong = goldsmith

[Mathelirium @mathelirium]

A Lens That Takes Derivatives

US Patent Basis: US8610839B2 - Optical Processing System for Computing Derivatives.

In a 4f Optical Processor, the first lens takes the incoming field and forms its Fourier spectrum. At that middle plane, a tiny optical mask multiplies the spectrum by iξ. This is the derivative operator written in Fourier language

u(x) -> U(ξ) -> iξU(ξ) -> ∂u/∂x

Then the second lens brings the field back to the real space. What comes out is no longer just a focused beam. It is the spatial derivative of the input field, computed by light as it propagates.

So, this is the serious promise of optical computing. A physical optical train can perform operations that usually live inside numerical code: differentiation, filtering, convolution, edge detection, correlation, and many other linear transforms.

[luthira @luthiraabeykoon]

We implemented @karpathy 's MicroGPT fully on FPGA fabric.

No GPU.
No PyTorch.
No CPU inference loop.

Just a transformer burned into hardware, generating 50,000+ tokens/sec.

The model is small, but the idea is not: inference does not have to live only in software 👇

[Kyle Chan @kyleichan]

Unitree founder Wang Xingxing:

In robotics, locomotion and basic motion is mostly solved. But grasping and manipulation—anything related to haptics—hasn’t been solved. That’s the key bottleneck preventing them from being deployed at scale in factories and homes.

He says that simulation is much faster for training but for manipulation tasks you still need real-world training data—for now.