A deep dive into linear and affine type systems, ownership semantics, lifetimes, and aliasing XOR mutability. How Rust verifies memory safety and delivers zero-cost abstractions at compile time, why cyclic graphs challenge ownership trees, and how these principles are mirrored in modern C# (.NET) with Span and ref struct.

Closing the Glance Rust series with a look back: how learning Rust's ownership model, enums, and compiler discipline quietly made me a better C# developer — and why the industry stopped treating Rust as speculative.

Cross the unmanaged border: Rust ownership, safety, FFI memory freeing traps, thread bounds, and resolving UI thread deadlocks inside Glance PDF.

In web development, memory is a request-scoped abstraction. On the desktop, state lives forever. Let's dive into .NET memory internals (Stack, Heap, LOH, and Pinning) through the lens of Glance PDF.

Building and training a 3-layer neural network from scratch in C# without ML frameworks or autograd. We derive the analytical backpropagation pass, utilize TensorPrimitives for SIMD-accelerated math, verify gradients numerically, and analyze MNIST performance with momentum SGD.

Learn how to build a self-hosted, high-performance facial recognition system in .NET using ONNX Runtime and the ArcFace model. We cover image preprocessing, cosine similarity, and database caching optimizations in production.

Dive deeper into compiler optimization, compilation models, and real-world projects. Learn advanced techniques like error recovery, optimization passes, and the difference between JIT and AoT compilation.

Learn compiler design fundamentals by building an algebraic expression compiler in F#. Understand tokenization, parsing, AST construction, semantic analysis, and code generation.

A from-scratch implementation of Llama 3.2 inference in F#: Grouped Query Attention, RoPE, SwiGLU, KV cache, and a Tiktoken tokenizer — running on .NET 10 with TorchSharp.

SIMD vectorization, zero-allocation backward pass, F# vs C# idioms, Adam optimizer, temperature sampling, and the limits of scalar autograd.