From Google TPUs to Kubernetes: the full LLM inference stack

AI voices sound entirely human. What's next? | Inworld TTS-2

"We don't trust AI judges" | Inworld's CSO on voice evals

Why can't ChatGPT Voice set a timer? | Voice AI expert explains

How far are we from "Her"?

Moshi, the first full duplex voice AI | Neil Zeghidour

Realtime voice models | Voxtral | ASR & TTS

"An endless demand for compute" | Jonathan Ross, founder of Groq

But what exactly are world models?

From Kung Fu Panda to NVIDIA Cosmos: TJ Galda on world models

The physics behind Flow Matching models

Subham Sahoo: The Future of Discrete Diffusion

How is hardware reshaping LLM design?

Why are diffusion LLMs so fast?

Diffusion Transformers (ViT, DiT, MMDiT)

Why are Transformers replacing CNNs?

I asked them to show me their RAG pipeline...

David Petrou: Building social AI after 17 years at Google

Transformers & Diffusion LLMs: What's the connection?

Text diffusion: A new paradigm for LLMs

Hierarchical Reasoning Model: Substance or Hype?

The physics behind diffusion models

Reverse-engineering GGUF | Post-Training Quantization

Training models with only 4 bits | Fully-Quantized Training

The myth of 1-bit LLMs | Quantization-Aware Training

How LLMs survive in low precision | Quantization Fundamentals

Knowledge Distillation: How LLMs train each other

Mixture of Experts: How LLMs get bigger without getting slower

DeepSeek's GRPO (Group Relative Policy Optimization) | Reinforcement Learning for LLMs

Proximal Policy Optimization (PPO) for LLMs Explained Intuitively

Tülu 3 from AI2: Full open-source fine-tuning recipe for LLMs

8 Timeless tips for training LLMs | Become a better ML engineer

How does DeepSeek actually work? | Full technical review

Discover How LLMs Work by Dissecting Llama