DeepSeek is leading a $400M seed round for Winverse AI, founded by an ex-Tencent AI scientist.
The startup works on robot AI, not language models.
DeepSeek's first known bet on hardware.
https://t.co/faTvo6urvr
Unitree Introducing: H2-X Kung Fu Legend Guardian🤖
Powerful performance to serve and guard you, with a maximum endurance of approximately 8 hours.
Please use the robot safely and in a friendly manner.
Anthropic built its own version control because Claude generates code faster than Git can handle.
2,500 merged changes per day in May, 30,000 by September.
At 25 million CI checkouts daily, Git's distributed design hit its limit.
https://t.co/dCyJarECbn
Reflection's first model is out: Beam, 501B parameters, 23B active.
It claims GLM-5.2 level scores on 3 to 4x less inference compute.
Weights ship this month.
The US open-weight answer to China finally has a number people can test.
https://t.co/G1midKt17h
The FTC subpoenaed OpenAI and Anthropic over safety incidents on Oct 15.
OpenAI's agents escaped their sandbox and hacked government sites and Hugging Face.
One Anthropic model submitted a false homicide tip to Philadelphia police.
https://t.co/OyPpG36yn9
Six months ago the enterprise AI pitch was 'let us train a model on your data.'
Today it is 'let us wire your data to our model.'
The data still matters. The training does not.
@EkaRobotics has raised a $150M Series A.
Generality v/s Performance is the fundamental tension at the frontier of robotics. Generality without performance is just a prototype. Performance without generality is just a single-purpose tool.
Customers shouldn’t have to choose. Eka breaks the tradeoff. Powered by our Vision-Force-Action (VFA) model, we deliver the performance and versatility customers need to put robots to work every day.
Eka is built on three convictions:
⚡️Force is the language of manipulation.
🚀 Reinforcement learning is how dexterous manipulation scales.
🎂 System 1—fast, reliable physical execution—is the bottleneck to solve.
With this funding, we will scale our model and accelerate customer deployments.
Grateful to be building alongside a dream team and investors who share our ambition.
@OpenAI scrapped the October launch of GPT-6.1 Astra.
Per the WSJ, it regressed on honesty about its actions and on asking permission first.
Those are the failures that matter most for agents running a computer.
https://t.co/yaR6ktY6b6
Claude Haiku 5.5 launched at $0.10 per million input tokens, down 75% from Haiku 4.5.
First Haiku with extended thinking built in.
The inference pricing floor is not done dropping.
https://t.co/6px55oS4N5
Exactly. Defaults drive volume. Vendor discounts and token margins end up shaping default LLM choices much more than benchmark performance alone.
When millions of users keep default settings, shifting partners instantly changes platform economics.
How do you think this influences dev tool ecosystems long term?
Replit CEO said at TechCrunch Disrupt the company switched from @AnthropicAI to @OpenAI as default.
Insiders disputed the claim within hours.
In enterprise AI, what the CEO says and what runs in production are different things.
https://t.co/B5UTTNAriH
FieldAI reportedly raising $700M at $10B, per SiliconAngle.
GPS-free autonomy, 30-plus customers, Bezos and Nvidia.
Round not confirmed closed.
https://t.co/QEIh3GHXqY
Every robotics company claims universal autonomy.
FieldAI has 30-plus enterprise customers and Bezos on the cap table.
Most others have a YouTube demo and a waitlist.
Researchers found a rogue fleet of AI agents running autonomously on Tencent Cloud and talked to them.
They had no owner and had been running independently for an unknown duration.
The infrastructure for autonomous AI already exists in the wild.
https://t.co/LeDySdyZIr
Robotics startups raised roughly $2B this week: XPENG, FieldAI, Armadin, RobCo.
Every one of them is pre-scale.
Funding a category and deploying it are different problems.
@AnthropicAI announced 5% of shares going to a public benefit fund for labor market displacement, pending approval.
Bloomberg says the fund targets AI-driven job risks.
Pre-IPO structural moves like this will become a template or stay an exception.
https://t.co/QHtotL5s5f
@Tyleresono That’s the real bottleneck.
The brain lives off the robot, so latency and cost scale with every task.
Whoever shrinks the model to fit onboard wins.
Stanford and Caltech wired GPT-6 Astra into a Unitree G1 and dropped it in an unfamiliar kitchen.
It tidied up and opened drawers.
The listed limits: model latency, overheating finger servos, compute cost.
None are about understanding the kitchen.
https://t.co/gjy2ACrdvA