Token Efficiency: Where the Tokens GoMay 14, 2026 Decode throughput is bandwidth divided by model size Single-token decode reads every weight once, so the spec sheet predicts tokens per second before you download anything. Measured against six Qwen models on an M3 Max.
llmai
Read article
May 5, 2026 DeepSeek V4 on Apple Silicon: Current Status What the V4-Pro and V4-Flash checkpoints weigh on disk and in memory, mapped to the unified-memory ceilings of current MacBook Pro and Mac Studio configurations.
llmaideepseekmlxapple-silicon
Read article
Token Efficiency: Where the Tokens GoMay 4, 2026 DeepSeek V4: Don't Look at What You Don't Need DeepSeek V4 reads a million tokens on roughly a quarter of V3.2's compute. It does this by selectively attending to the parts of context the prompt asks about, the same way humans skim a long book.
llmaiattentiondeepseek
Read article
AI Goban/Baduk Board DetectorApril 7, 2026 Converting Custom Annotations to YOLO Keypoint Format My corner annotator outputs JSON. YOLO wants normalised bounding boxes with keypoints in a specific folder structure. Bridging the two formats.
computer-visionpythongoai
Read article
AI Goban/Baduk Board DetectorApril 6, 2026 Corner Annotation, Board Scraping, and Training YOLO No off-the-shelf model could find board corners reliably. I built a manual annotator, scraped training data, and trained a YOLO model on the results.
computer-visionpythongoaigpu
Read article