AI 热榜

DeepSeek releases DSpark - 50%-600% faster spec decoding vs MTP

DeepSeek just released DSpark for V4 Flash & Pro, a new speculative decoding

Chinese AI startup MiniMax plans to open-source a 2.7 trillion parameter model later this year

Meta tests always-on AI glasses that capture your entire day

OpenAI's GPT-5.6 launches Thursday after a delay forced by the U.S. government

Hot French startup ZML releases free product to speed inference across lots of AI chips

AI chip maker SambaNova raises $1B at $11B valuation, 5 months after last mega round

Hackers can use 9 of the most popular AI tools to assemble massive botnets

GitLost: We Tricked GitHub's AI Agent into Leaking Private Repos

Ant Group’s Robbyant Open-Sources LingBot-Vision: A 1B Boundary-Centric Vision Foundation Model for Dense Spatial Perception

Why the rise of open source AI isn’t hurting Anthropic … yet

Deepseek drops another HUGE breakthrough - DSpark. Waaay faster than MTP [Video explaining it]

We charge $10k a week to delete AI-generated code

Former OpenAI exec Kevin Weil is now on the board of Stoke Space

From Hugging Face to Amazon SageMaker Studio in one click

Microsoft joins AI cost-cutting trend by relying more on its own models

Discord admits AI moderation bug wrongfully banned users over harmless images

We are excited to launch muse Image, the first media generation model built by MSL. Muse image can use reasoning, refinement, and tools to improve pre...

and that's still the old Fable Anthropic is in their own tier right now

I was expecting Kimi K3 to be the first 2-3T model I guess MiniMax M3 Pro at 2.7T will also do

looks like they'll exceed 11000 TWh this year at this rate maybe 8 MWh per capita About Austrian levels, long past EU average 1-2 more years to exceed...

yes they are preparing for the GPT-5.6 launch tomorrow max and ultra reasoning incoming

From SemiAnalysis: Anthropic Q3 profit at over $1B Their base-case is that Anthropic will be the first 6 trillion dollar company

"I haven’t needed to check the code I’ve written in two months" looks like OpenAI had the model for a while

Wait: He has already been testing 5.6 for months? Plural: months, not singular: month. That means 5.6 had already finished training when Mythos and Fa...

Beauty GLM-Image was trained on Ascend 910s, and now runs on Apple

AI native products: - minimize cleverness - maximize configurability - API first Basically the opposite of what you want in a human native product

Codex for creating the first video of a black hole:

I spent an hour making a chart, pouring every bit of taste and discretion into it. I thought I nailed it. I showed it to Claude for a spellcheck. And ...

This is what latency optimization looks like below the API 👇 Together ATLAS, NVIDIA Blackwell, CUDA, TensorRT-LLM, Dynamo, and custom kernels all w...

check it out

> Unitree's supply chain for joint motors extends little beyond raw materials such as copper wire and magnets this is why Unitree will remain formidab...

Day 0 support coming for Grok 4.5 as well, of course!

Next up at the booth today at 12:30pm, see how Gemini collaborates with artists to design complex origami. Join Tom Zahavy, Xidong Feng, and Shaobo Ho...

Why are they forecasting without Fable?

codex can pursue goals for longer durations than some of you

Vibes are off like you wouldn't believe Feels like *everyone* is going to lose at this rate Russia, Ukraine, Argentina, Iran, USA, China, open source,...

“Researchers from <brand name company/school> just **proved** that AI has a blind spot 60% of the time!” And its a prompt engineering paper

we don't invite amdahl to parties but he shows up anyway

Its been real

UI still needs some love but Fable has been a beast at taming my AI chat archives and distilling them into something actionable.

> but by 2009 China already had almost as much infrastructure as it could productively absorb ahaha. by 2025, China still had places near the capital ...

That said, this item in the 15th Five Year Plan is copium fuel

Nice detail:

I released sqlite-utils 4.0, the 124th release but the first major version bump since 3.0 back in 2020 I managed to keep things backwards-compatible a...

True From what I know, we can write Google off if even *this* is a cursed-bloodline embarrassment due to half-assed data pipeline. People will flee. G...

It's been an amazing six years. I'm thrilled that Sayash will be starting his own group and soaring to new heights. This is the part where I would tal...

Amazing work! If anyone has some spare compute, send it his way so he can finish this up.

My standards for V4-Pro-Official are: match GLM 5.2 on net. I'll accept something like slightly worse DeepSWE for slightly better CritPT/FrontierMath,...

One of the weirdest goals one could set, but infinitely valuable if it could be achieved: deliberate design of p-zombies.

IMO the greatest philosophy is Science Fiction - ideas explored through the prism of changes in our technological reality, breaking free from default ...