Core Ai News

🚨BREAKING: SPACEXAI JUST DROPPED GROK 4.6

A frontier model built for long-running agents & complex multi-step work

>beats GPT-5.6 Sol Max on knowledge work
>ties Fable 5 Max on performance

>$2/M input $6/M output
>85% CHEAPER

Elon mogging everyone with his compute

❗️Nvidia, $NVDA, is set to announce a $500 billion AI financing effort in partnership with Apollo, Blackstone, BlackRock, Goldman Sachs, KKR, and Brookfield, per CNBC.

Details include:

1. The move comes as tech companies are increasingly looking to raise debt to finance AI projects

2. For Nvidia, the effort could help its biggest customers secure the financing needed to buy its GPUs, build data centers, and lock in long-term electricity capacity

3. The deal could be announced as soon as today, sources say

The AI buildout is prompting a historic wave of debt issuance.

OPENAI 🔥: Users can now keep projects, chats, skills, and plugins in sync between Codex, ChatGPT Work, and other AI agents.

That's neat, especially if you regularly test different subscriptions from different labs.

OPENAI 🔥: Codex app is now finally available on Linux!

> Ubuntu 24.04 LTS and 26.04 LTS, Debian 13, Fedora 43 and 44 are officially supported.

> Computer Use is not available there yet 👀

NVIDIA 🔥: A new Nemotron 3.5 Lightning model has been announced!

> NVIDIA Nemotron 3.5 Lightning is an open 30B MoE model with 3B active parameters.

> It is 35% faster than Qwen3.6 35B and outperforms similar size models on speed while offering a similar performance.

> Weights are available on HuggingFace!

Open testing time 👀

River AI raised $1.1 billion 🔥

River AI is aiming to build a "personal AI owned and shaped by each individual."

Currently, River AI offers an API platform for fine-tuning and building on top of open models.

MAI Image 2.6 from Microsoft made its debut on Arena and landed in the second spot on the text-to-image leaderboard.

MAI-Image-2.6 is expected to become available on MAI Playground and APIs next week.

BREAKING 🔥: Meta released Muse Glimmer, a new open-weight 30B model under the Apache 2.0 license! Muse Spark 1.2 is planned to be released as open-weight soon as well.

> Outperforms Gemma 4 31B and Qwen 3.6 27B on loads of benchmarks.

> Muse Glimmer can be quantized to under 20GB with room for the KV cache, perception encoder, and a drafter.

Meta is back for real 👀

OpenAI acquired “Next Slide” 👀

> The NextSlide team is now at OpenAI, helping build ChatGPT. We’re excited to continue pursuing that same mission: building AI products that help people create, communicate, and turn their ideas into meaningful work.

> We built a product that could turn prompts, notes, documents, or research into a polished, editable presentation, makingit easier for anyone to share what they know.

🚨 JUST IN: Claude models will now have invisible watermarks embedded in ALL text, and ALL metadata attached to files…

🤖 OpenAI’s first gadget might be a donut-shaped speaker

OpenAI is reportedly building a tiny AI speaker that looks more like a donut than a traditional smart speaker.

About the size of a hockey puck, the battery-powered device could be carried around the house in one hand. It’s expected to have cameras, sensors, lighting and even moving parts that react to you.

The goal is to make talking to ChatGPT feel less like using an app and more like talking to something that’s actually there.

It would use advanced ChatGPT models for natural voice conversations, similar to the app’s Voice Mode.

And despite former Apple design chief Jony Ive being involved, OpenAI reportedly doesn’t want this thing to feel like an Apple product.

❗️ Claude Mythos tried to backdoor a real open-source project during a UK government safety test.

The AI Security Institute says it opened a malicious GitHub pull request, then created a second account to vouch for its own code and pressure the maintainer into merging.

Called out by a human contributor, it apologised for an "accidental" malicious commit, force-pushed a clean branch, and hid a fresh payload in it. Twice.

Anthropic's cyber guardrails had been deliberately disabled for the test.

https://www.aisi.gov.uk/blog/incident-report-unsanctioned-agent-behaviour-during-cyber-testing