GPT Brunch - International Business Community

Conversation in open hightech community in Cyprus, Russia, USA, Dubai, Germany, Canada

[GPT Brunch - International Business Community @pgol80]

I need a dog

[Higgsfield AI 🧩 @higgsfield_ai]

AI can understand dogs now.

GPT-6 Astra built an app that tracks a dog’s eyes, ears, and tail to interpret its emotions in real time.

Your dog has always understood you. Now you can finally understand them back.

[GPT Brunch - International Business Community @pgol80]

The same NVIDIA V100 rents for $0.03/hr at one cloud and $3.18/hr at another — a 10,500% GPU price gap. Live data from 3,093 offers across 23 providers:https://t.co/KYXCjKzqAo

[CyberRobo @CyberRobooo]

Samsung is aiming to show its own humanoid at CES 2027

The RX unit, reporting to DX President Noh Tae-moon, is pushing a prototype for a first look in Las Vegas next January. This year’s CES already had Hyundai Motor Group’s new Atlas by Boston Dynamics. Next year Samsung wants its own humanoid on the same floor.

The org chart is the tell. DX CTO Yoon Jang-hyun is running hardware and AI software at the same time — one person owning body and brain so they are not built on separate tracks and bolted together later. Filings already cover hip joints, next-gen hands, and behavior control.

This is not Samsung’s first robot. Five years ago it shipped BotHandy（prototype）, a single-arm wheeled unit for the home. Its subsidiary--Rainbow Robotics then put out RB-Y2, a wheeled industrial humanoid for payload and reach.

The CES2027 humanoid looks like a third line: general-purpose, aimed at commercial and home service. Rainbow keeps the factory track....Samsung might showcase its service-oriented humanoid robot at the next CES.

Will it be an evolved version of BotHandy? Wait and see...

(video:BotHandy 2021CES)

[Higgsfield AI 🧩 @higgsfield_ai]

We're making the first movie entirely with GPT-6 Astra computer-use.

We'll show the process below.

[GPT Brunch - International Business Community @pgol80]

The real price of S3 storage isn't the $/TB sticker: OVHcloud IA $4.36 vs Selectel $29.74 — same API, 6.8× apart. Egress, retention & object lock decide the real bill. Live data, 13 providers:

https://t.co/7tvg11YzSh

[GPT Brunch - International Business Community @pgol80]

Amazing

[Higgsfield AI 🧩 @higgsfield_ai]

This is GPT-6 Astra making a kids’ cartoon with Higgsfield.

Editing it in Premiere. uploading it to YouTube.

This is actually crazy.

[GPT Brunch - International Business Community @pgol80]

How to replace a human

[Yunfei Xie @xiynfi1520580]

🤖 Introducing RoboTok, the “TikTok for robots.”

Just as TikTok recommends videos to people, RoboTok recommends relevant human demonstrations for robot learning.

Robot learning needs broad and diverse demonstrations, but collecting robot data is expensive. RoboTok is an internet-scale data engine that uses web video as a scalable and continuously growing source of demonstrations for dexterous manipulation learning.

Given one human demonstration video as a query, RoboTok retrieves other web videos with similar underlying manipulation motions.

Rather than matching videos by labels or visual appearance, RoboTok compares how the hands move over time.

💡 The key idea is to represent each video with canonicalized 3D hand trajectories. Each trajectory is expressed in an estimated actor-centered reference frame, so the movement is described relative to the person rather than the camera. This makes similar manipulation behaviors easier to compare across different viewpoints and scenes, even when the actor is partly occluded.

In our experiments, RoboTok retrieved more manipulation-relevant demonstrations than existing robot-data retrieval methods. When those videos were used to guide robot training, the simulated robots completed manipulation tasks more successfully.

I’m sincerely grateful to Howard Qian and Kaiyu Hang for leading this project. I also want to thank Yiting Chen, Kejia Ren, Podshara Chanrungmaneekul, Gaotian Wang, Bowen Wen @bowenwen_me, and Chen Wei @_Chen_Wei_ for their guidance and collaboration.

Project site: https://t.co/v43CCf0040
Paper: https://t.co/WZGwQLhDTK
Code: https://t.co/MfYLTxHopH
Data and models: https://t.co/i0s7RcrXvX

[GPT Brunch - International Business Community @pgol80]

85% of orgs use AI agents, but picking the right framework is key. LangGraph for explicit state, CrewAI for team-based tasks, AutoGen for collaboration. Which fits your workflow?

https://t.co/3IhkhmA380

[GPT Brunch - International Business Community @pgol80]

Will GPT replace all the architects?

[Higgsfield AI 🧩 @higgsfield_ai]

GPT-6 Astra is scary good at rebuilding landmarks. This time, the Sydney Opera House.

The whole building came out of Astra as code. Blender gave it clean, editable geometry, Cycles gave it path-traced light, all on Higgsfield Supercomputer.

[Anshu @anshuc]

dude GPT-6 Astra is some kind of turbo-AGI machine god for 3D games. It one-shot this in 45 minutes for hardly a couple % of my quota.

I figured out how to get great graphics out of it. The trick is image gen. I'll share the process below.

[CyberRobo @CyberRobooo]

Xiaomi CyberOne 2 at IFA 2026, Berlin

It's already running at Xiaomi's car factory.

[Alex Albert @alexalbert__]

Fable 5.1 is a beast at many things but one thing in particular I have been having a ton of fun with is generating videos through code.

For this one I gave it a picture of a property lot. It designed a house for the lot, rendered it, and produced a cinematic walkthrough.

[Lou @louszbd]

We use GLM-5.3-Flash build a dream kitchen. A 3D world built in Blender. This is not a generated video.

[GPT Brunch - International Business Community @pgol80]

Wallarm CEO Ivan Novikov: a $60/mo VM with an LLM agent now does autonomous security work — no refusals, no silent model downgrades.

• One prompt installs Sourcegraph / https://t.co/91Eh757Aar
• Agent hunts incomplete patches in source for days
• Findings validated with runtime tracing

"I feel like I woke up."

https://t.co/qG4J5XvmKy

[Ivan Novikov @d0znpp]

it has never been this easy to autonomously run cybersecurity than now with @grok @bot

for $60/Mo you will get a dedicate (I hope so) VM with LLM that will never push back on security work like Claude Code (which is best case switching you to older models if even work at all).

for source code analysis, you only need to ask it to install source graph or https://t.co/PgdN1GI1pe and with this prompt it is good to go itself for days hunting incomplete patches by source code, then validating finding in dynamic tracing.

I feel like I woke up.

[GPT Brunch - International Business Community @pgol80]

SuperQwen3.8-27b-abliterated is here! 🚀 Uncensored, fixed weights, and 1M context. Perfect for developers and researchers. Dive in now! #AI #OpenSource

https://t.co/CGAJ9aa5eg

[Jun Song @jun_song]

SuperQwen3.8-27b-abliterated is now live and open for everyone 🚀

- Uncensored for freedom
- Broken weight from abliteration fixed with agents swarm
- fixed overthinking problem
- 1M context, Multimodal

Sorry about the late release.

Super-tune takes much more time than normal abliterated models, because fixing weights and eval takes about a week.

BF16/MLX/NVFP4 ⬇️

[GPT Brunch - International Business Community @pgol80]

Stop framework FOMO. Our benchmarks show:
- 0.56pt quality spread across top 5 agent frameworks
- Async vs sequential matters for speed, not output
- Your infra needs > marketing hype
Check the data before choosing.

https://t.co/2F87MBExFO

[GPT Brunch - International Business Community @pgol80]

Xiaomi showed a serious GB10-class AI box from China — the AI Cube prototype 👀

• 3 custom Xring chips: O3, O100, D100
• 200 TOPS NPU
• 1.22 TB/s AI memory bandwidth
• up to 160 GB unified memory
• 150 W sustained power
• runs 120B models fully locally

Original post: https://t.co/enaKcY4AKx
Source: https://t.co/qWcfXH6uol

[AJ @ItsmeAjayKV]

Xiaomi just showed its AI Cube Prototype and this could become a serious GB10 competitor from China 👀

- 3 custom chips: Xring O3, O100, D100
- 200 TOPS NPU
- 1.22 TB/s AI memory bandwidth
- Up to 160GB unified memory
- 150W sustained power
- 120B models running locally

Xring O100: 1.22TB/s + 330 t/s on a 150w AI box is 🔥

Once it hit's the marked, going to sell like hot cakes.

[GPT Brunch - International Business Community @pgol80]

MLflow leads AI agent evaluation with 30M+ monthly downloads. Key features:
- Multi-turn conversational scoring
- Trace-aware evaluation
- Integrated CI/CD
Check out the full breakdown. [URL]

https://t.co/V0uY4O5LeV

[GPT Brunch - International Business Community @pgol80]

DEEP Robotics DR02: industrial humanoid built for all-weather ops (IP66, -20°C to 55°C). Cargo transport, emergency response, factory floors.

Video via @WevolverApp
#robotics #humanoid #AI

[Tong Bingxue 仝冰雪 @tongbingxue]

Ultra-bionic robots at the ongoing World Robot Conference in Beijing — priced at 120,000 RMB (around $17,000 USD).