4d ago · 9 min read · Instead of requiring the full compute footprint of a 400B-parameter model at every step, Qwen3.5 dynamically activates only a subset of its parameters. This allows developers to access large-model int
Join discussion
Feb 3 · 21 min read · It's Been a While Sorry for being away for so long. I've been head-down strengthening my software engineering foundation—consuming rather than producing. After a while, I started feeling this pull to balance things out by building again. I just finis...
Join discussion
Dec 20, 2025 · 5 min read · Most data in the world today is visual. We take photos of receipts, scan documents into PDFs, record meetings on video, and capture dashboards as screenshots. All of these visual formats are full of valuable information. But turning that visual infor...
Join discussion
Nov 17, 2025 · 13 min read · Why Pixels Are the New Tokens: A Deep Dive into DeepSeek's Genius OCR Remember when you glanced at a poster across the street and instantly understood what it said? Your brain didn't process every single letter individually. It compressed the visual ...
Join discussion
Oct 30, 2025 · 2 min read · 📖 What Are Multimodal Models? Multimodal models are AI systems that can see, hear, and understand multiple data types — such as text, images, audio, and even video — all at once.Unlike traditional text-only models, multimodal AIs like GPT-4V, Gemini...
Join discussion
May 25, 2025 · 2 min read · I got a free ticket to the Embedded Vision Summit 2025 in Santa Clara last week, and figured I'd share some notes while they’re still fresh in my head. The main theme? Vision Language Models (VLMs) and multi-modality. If you haven’t heard of VLMs, th...
Join discussion
May 18, 2025 · 10 min read · Artificial intelligence (AI) has changed the way we live and work. Today, many people use AI tools to make everyday tasks easier. Some even rely on them for facts or to help make important decisions. I may or may not have used AI myself while writing...
Join discussion
Nov 19, 2024 · 3 min read · Scientific strides often entail wading through a sea of technical complexity to unveil innovations with the potential to transform industries. One such advancement comes from a notable paper that introduces BEND-VLM, a method designed to improve fair...
Join discussion
Oct 18, 2024 · 11 min read · Llama 3.2, a powerful multimodal large language model (LLM) from Meta AI, has recently been released, pushing the boundaries of AI capabilities by enabling machines to understand both visual and textual information. While this pre-trained model is im...
Llin commented