Core Ai News
2.52K subscribers
1.48K photos
220 videos
1 file
346 links
• Daily AI updates
• Latest breakthroughs
• Machine learning & deep learning
• AGI & LLM news
• AI research & trends

Other channels
- @CorePrompts
- @CoreAgents
- @Coretutorial
- @CoreUtil

Run by a Human
Download Telegram
ICYMI 👀: Z AI aims to reach Fable 5 level already this year.

xAI wen?
1👾5
BREAKING: The NSA confirms Mythos “broke into almost all of our classified systems, not in weeks, but in hours”
📱 Users can now search for Imagine images and videos on Grok!

It works like a proper image search but scoped to your Imagine creations. I wish we have it across the whole set of published Grok images eventually - that would be huge!

📱 Meta AI is getting a new Artifacts tab on the web. All the presentations, docs, web pages and other creations would be stored over there.
Please open Telegram to view this post
VIEW IN TELEGRAM
👾37
OpenAI just hired one of the architects of modern AI.

Noam Shazeer, a legendary AI researcher and one of the key minds behind the transformer revolution, has reportedly joined OpenAI to lead AI architecture research.

If his name sounds unfamiliar, his work definitely isn’t. Shazeer co-authored the 2017 paper Attention Is All You Need, which introduced transformers, the breakthrough that powers ChatGPT, Gemini, Claude, and nearly every major AI model today.

His résumé is stacked: multi-head attention, Mixture-of-Experts (MoE), and T5 all trace back to ideas he helped create.

After more than two decades at Google, Shazeer left in 2021 to found Character.AI. Google later paid a staggering $2.7 billion to bring him and the startup back, installing him as a key leader behind Gemini.

Now, one of Google’s most influential AI minds is heading to OpenAI.
👾7
Media is too big
VIEW IN TELEGRAM
‼️U.S. Commerce Secretary Lutnick is concerned one of ASML's lithography chipmaking machines may have ended up in China.

ASML's extreme ultraviolet lithography machines cost around $370 million and take multiple 747s to transport.

ASML denies the allegation, saying it has never shipped an EUV machine to China.
👾38
‼️🇳🇴 Norway bans the use of AI in primary schools due to declining school results.

A new national recommendation says pupils should largely not use AI in schoolwork. The government's reasoning is sequencing: kids need to master reading, writing and arithmetic first, and it says research shows uncritical use of generative AI lets them skip the hard parts of learning.

The backdrop is a years-long slide in basic skills. The government cites data showing one in four Norwegian pupils now reads below the minimum level, and says it won't repeat the mistake of rushing digital devices onto young children.
👾39👍3
🚨MICRON ANNOUNCES STRATEGIC DEAL WITH ANTHROPIC

- Multi-year HBM, DRAM and SSD agreement
- Co-designing memory + storage around Claude’s workloads
- Claude deployed across Micron
- Micron invested in Anthropic Series H

Micron is now Anthropic’s investor, customer, partner, AND supplier.
👾38
NEW: SpaceX discloses a $100,800,000,000.00 cash balance.
👾38
🚨JAPANESE AI STARTUP JUST MATCHED CLAUDE FABLE 5 AND MYTHOS PERFORMANCE.

Japanese AI lab just launched Fugu, a model trained to command other models.

Sakana AI, a Tokyo-based AI startup, was co-founded by researchers including one of the authors of the original Transformer paper, the blueprint that basically every AI model today is built on.

Sakana Fugu is a multi-agent AI system that feels like a single model, automatically orchestrating specialized models behind one API endpoint.

Fugu Ultra is claimed to match Anthropic's Fable 5 & Mythos Preview on the hardest engineering/science/reasoning benchmarks — and to beat Gemini 3.1 Pro, Opus 4.8 & GPT-5.5 on tasks like AutoResearch, mechanical design & financial forecasting.

Everyone's racing to scale ONE giant model but Fugu flips it: an LLM that orchestrates a pool of the world's best models (choosing who does what, delegating, verifying and merging it into one answer), even calling itself recursively.

Think of it as a conductor. You send one request to one API, and Fugu figures out the move: answer it solo, or assemble a squad of expert models and run the whole thing for you.

Sakana AI launched two models:
Fugu - fast, low-latency, everyday coding/chat
Fugu Ultra - max quality on hard, multi-step problems (AI research, paper reproduction, cybersecurity, patent search)

Export controls are hitting frontier models now, but because Fugu's agent pool is swappable, if a provider restricts access, it routes around it.

Both Fugu and Fugu Ultra can be accessed through a unified API, with subscription plans for everyday users and pay-as-you-go pricing designed for high-volume and enterprise workloads.
👾41
‼️🚨 BREAKING: OpenAI just launched a new cyber model that beats Mythos on CyberGym, a benchmark for finding real software bugs.

The real story: OpenAI just upgraded the permissive, exploit-capable cyber model it already gives "verified defenders," and the new version nearly doubles GPT-5.5 at turning known bugs into working exploits (39.5% vs 25.95%), still gated behind monitoring and government access deals.
👾3
BYTEDANCE 🔥: Seedance 2.5 has been officially announced, along with an updated Seedance 2.0.

- Seedance 2.0 now supports 4k output
- Seedance 2.5 will be able to generate 30-second videos in one go
- ByteDance also announced a new AI copyright commercialization platform
👾38
JUST IN: OpenAI’s new GPT-5.5-Cyber model beat Mythos 5 on the CyberGym benchmark.
👾39
🤖 It looks like we’re getting a whole range of new GPT models this Thursday:

• GPT-5.6
• 5.6 Pro
• And a new bidirectional voice model.

@CoreAti
Mistral AI launched OCR 4 👀

> Win rates averaging 72%, alongside the top overall score on OlmOCRBench (85.20).
> Alongside the extracted text, OCR 4 returns bounding boxes, typed-block classification, and inline confidence scores.
> OCR 4 is an ingestion component of Search Toolkit, Mistral's open-source, composable search framework.
> Support for 170 languages across 10 language groups.
> OCR 4 is compact enough to run in a single container.
“A man who has not hit his Claude limit by noon has wasted his morning.”

-Socrates
2🤣193💯2💅111
🚨BREAKING: OpenAI and Broadcom announced their first AI Chip, named “Jalapeño”! designed specifically for LLM workloads.

- Built from scratch for LLM inference, not general-purpose compute
- Early tests show much better performance per watt than current leading AI accelerators
- Developed in just 9 months, with OpenAI models helping speed up chip design
- Part of OpenAI’s strategy to own the full stack from models to hardware
- Co-built with Broadcom and Celestica for large-scale deployment
- Planned rollout at gigawatt scale starting in 2026
- Goal is to make AI products like ChatGPT and Codex faster, cheaper, and more reliable