Hello, the number of times car brands are cited when asking for EV recommendations to different AI models.
RESEARCH
-
DeepSeek Achieves 50x Attention Efficiency Breakthrough
By
–
DeepSeek casually unlocked 50x attention efficiency in ~1 year > MLA is ~5.6x faster than MHA
> DSA is 9x faster than MLA never doubted you, you big beautiful whale -

What Should Replace MMLU for AI Model Evaluation?
By
–
Final update! MMLU is saturated and has become (rightfully) less popular. What should replace it? – Other knowledge evals like GPQA, MMLU-Pro
– Code evals like LiveCodeBench
– Agentic evals like BFCL
– Other? -
GATO Project: Building a Generalist Agent Model with GPT
By
–
Thank you … this certainly was and still is one of my favourite projects. The moment @scott_e_reed and I saw GPT we started working on Generalist AgenT One (GATO). We strongly believed a model could do anything a human could do from motor control, to perception, generation,
-

DeepSeek-V3.2-Exp: 50% Cheaper, Better Search
By
–



DeepSeek released DeepSeek-V3.2-Exp build on top of previously released V3.1-Terminus model. It is 50% cheaper and slightly better at search benchmarks. Deep dumping
-

AI Evaluation Shift: From Recognition Tasks to Economic Value Metrics
By
–
Hace 10 años evaluábamos a la IA por su capacidad de leer textos sencillos o reconocer perros/gatos en imágenes. En 2025, con benchmarks como el nuevo GDPval, se evalúa por su capacidad de resolver tareas económicamente valiosas que contribuyan al PIB. ¡El salto de una década!
-
Newton’s Laws: The Ultimate AI Model Without Training Data
By
–
My model needs zero training data, has only a few parameters, and can predict all future inputs from just the current one. It’s called Newton’s laws.
-

Rapid AI Progress Toward AGI by End of 2026
By
–
In 2021, AI could barely handle 10 minutes.
In 2025, it executes 2 hours of tasks autonomously.
Mid-2026: 8 hours.
End of 2026: parity with human experts across dozens of professions.
AGI is arriving faster than expected. -
VLQ: Europe’s Quantum Computing Initiative for Research Innovation
By
–
VLQ: Europe’s Next Leap Towards Quantum-Powered Research and Innovation – EuroHPC JU https://
eurohpc-ju.europa.eu/vlq-europes-ne
xt-leap-towards-quantum-powered-research-and-innovation-2025-09-23_en
…
#digitaleu #quantum #Innovation #technology @DigitalEU @ArturHabant @elaniazito @BetaMoroney @mvollmer1 @mikeflache @CurieuxExplorer @Shi4Tech @AnthonyRochand @lyakovet -
AI Progress Models: Beyond Linear Log-Scale Assumptions
By
–
It’s remarkable that so many of the people training the most complex models ever can’t think of a better model for their progress than an endless straight line on a log plot.