AI is in its self-improvement era: OpenAI says its new coding model helped to build itself

On Thursday afternoon, OpenAI released a new cutting-edge coding model that the company said assisted in its own creation. “GPT-5.3-Codex is our first model that was instrumental in creating itself,” ...

New AI model enables native speakers and foreign learners to read undiacritized Arabic texts with greater fluency

Reading an Arabic newspaper, a book, or academic prose fluently, whether digital or in print, remains challenging for many ...

Want local vibe coding? This AI stack replaces Claude Code and Codex - and it's free

And then there's agentic AI coding. When a tool can help you do four years of product development in four days, the impact is world-changing. While vibe coding has its detractors (for good reason), AI ...
Microsoft

Detecting backdoored language models at scale

Learn how Microsoft research uncovers backdoor risks in language models and introduces a practical scanner to detect tampering and strengthen AI security.

Cracking the rules of gene regulation with experimental elegance and AI

Gene regulation is far more predictable than previously believed, scientists conclude after developing the deep learning ...

Qwen3-Coder-Next offers vibe coders a powerful open source, ultra-sparse model with 10x higher throughput for repo tasks

On SWE-Bench Verified, the model achieved a score of 70.6%. This performance is notably competitive when placed alongside ...

Are Large Language Models A Dead End Or Simply Incomplete?

Once a model is deployed, its internal structure is effectively frozen. Any real learning happens elsewhere: through retraining cycles, fine-tuning jobs or external memory systems layered on top. The ...

I hacked my own computer using OpenClaw and it was terrifyingly easy

Agentic AI tools like OpenClaw promise powerful automation, but a single email was enough to hijack my dangerously obedient ...

I tested local AI on my M1 Mac, expecting magic - and got a reality check instead

At random, I chose glm-4.7-flash, from the Chinese AI startup Z.ai. Weighing in at 30 billion "parameters," or neural weights, GLM-4.7-flash would be a "small" large language model by today's ...
The Lancet

CARDBiomedBench: a benchmark for evaluating the performance of large language models in biomedical research

Although large language models (LLMs) have the potential to transform biomedical research, their ability to reason accurately across complex, data-rich domains remains unproven. To address this ...

Arcee's U.S.-made, open source Trinity Large and 10T-checkpoint offer rare look at raw model intelligence

San Francisco-based AI lab Arcee made waves last year for being one of the only U.S. companies to train large language models (LLMs) from scratch and release them under open or partially open source ...

Poetiq nabs $45.8M in seed funding for its LLM-enhancing ‘meta-system’

Poetiq was launched last year by former Google DeepMind researchers Shumeet Baluja and Ian Fischer. Baluja established the Alphabet unit’s computer vision group, while Fischer helped develop some of ...