Search the public knowledge base.
42 results

Type hello, world! into an LLM. To you, that string is already meaningful: two familiar words, a comma, a space, and an exclamation ...

Take a simple completion: “Alice gave Bob the book because wanted it.” Suppose the model assigns a high target logit $y$ to the corr...

A robot is about to push a puck across a table. There is a target marked a little farther away, and beyond it, the edge. A gentle pu...

Suppose you want to train a model with a matrix optimizer such as Muon. The optimizer does not think of a weight matrix as a bag of ...

A plugin registers commands, subscribes to events, allocates resources, and changes shared state. Loading is easy because the host c...

Imagine an LLM solving a math problem. It samples a full reasoning trace, you run a verifier, and the verifier returns a number: per...

Imagine training a robot arm from a replay buffer that contains plenty of examples of ordinary motions—move left, move right, lower ...

A diagram-rich generated explanation from the public library.

A diagram-rich generated explanation from the public library.

A diagram-rich generated explanation from the public library.

A diagram-rich generated explanation from the public library.

Backend Engineering from first principles

A diagram-rich generated explanation from the public library.

A diagram-rich generated explanation from the public library.

A diagram-rich generated explanation from the public library.

AI guardrails and saftey

A diagram-rich generated explanation from the public library.

Harness engineering

A diagram-rich generated explanation from the public library.

A diagram-rich generated explanation from the public library.

KV Caching

Speculative Decoding in LLM's

A diagram-rich generated explanation from the public library.

A diagram-rich generated explanation from the public library.

A diagram-rich generated explanation from the public library.

AlphaGo paper combining deep neural networks with tree search for superhuman Go play

VLA (Vision Language Action Models)

MoE architecture for efficient LLM scaling via specialized experts

RL framework that approximates maximum likelihood for binary-outcome tasks.

GRPO and RL for LLM's

Explain Kubernetes architecture including networking in depth

Stable JEPA-based world model that learns and plans from raw pixels.

Language models that recursively refine or compose intermediate reasoning/representations.

Vector Embeddings and Vector Databases

Mamba-3: Improved Sequence Modeling using State Space Principles

World Models

Self-supervised vision model for learning image representations.

Transformers

Diffusion and flow-matching

LORA Fine-tuning (Low rank adaption)

Variational auto-encoders

policy gradient methods