Tried Qwen3.6-27B on both Ollama and MLX. Ollama couldn't load the VL-projector-embedded GGUF, MLX ran it at 11 tok/s. On the side, running 35B-A3B under MLX was roughly 2× faster than the Ollama GGUF. Also had both models build a BBS to gauge intent handling.
A hub for the 5-article series that organizes math symbols in AI and LLM articles for reading, not solving. Covers equations, vectors and matrices, probability and statistics, derivatives, and gradient descent with backprop, plus a reading-order guide for different backgrounds.
Gradient descent, SGD and Adam, backpropagation, vanishing/exploding gradients with residual connections, and learning rate schedules — organized around what each piece is doing at a high level. The goal is reading training logs and model card numbers, not computing anything.
A minimum set of calculus for reading AI and LLM articles — d/dx, e, the chain rule, partial derivatives, and gradients. Focus on what the symbols are doing, not on solving the formulas.
A minimum set of probability and statistics for reading AI and LLM articles — conditional probability, cross-entropy, perplexity, and temperature are the main ones; rigorous Bayes and MLE derivations stay out of scope.
A browser tool that reads MusicXML and returns fingerings tuned to your hand size and biomechanical constraints. Walks through the backtracking cost minimization, the actual weight values, the academic lineage since Parncutt 1997, and why the same framework generalizes to guitar.
iPhone 17's HEIC adds new brand identifiers to the ftyp box, pushing it past exifr's hard-coded 50-byte guard. Here's a dynamic-import fallback to ExifReader, plus Null Island filtering and iloc pre-inspection to harden browser-only photo tools.
A minimum set of vectors and matrices for reading AI and LLM articles — the dot product and matrix product are the main two; determinants, inverses, and eigenvalues stay out of scope.
A regression in cryptographic signature validation introduced a CVSS 9.1 flaw into .NET 10.0. The Data Protection API implemented HMAC verification incompletely, opening the door to padding oracle attacks and forged authentication tokens.
A minimum set of math for reading AI, LLM, and image-generation articles — the aim isn't to derive anything, just to recognize weighted sums, S-curves, probabilities, and the 'nudge toward the answer' step of training.
Hands-on Qwen3.6-35B-A3B (23GB 4bit GGUF) on M1 Max 64GB via Ollama 0.20.6. Generation speed stays at 27 tok/s — same as Qwen3.5-35B-A3B — but the same prompt produces 13× more thinking tokens. Multi-turn behavior, persona handling, and a three-tier NSFW probe included.
Hands-on run of trellis-mac (the CUDA-free port of TRELLIS.2) on M1 Max 64GB. Setup via uv with PyTorch 2.11.0 MPS, applied mps_compat.py patches, and recorded actual generation time vs the M4 Pro 24GB 3.5-minute reference, plus where the bottlenecks land on Apple Silicon.
