Research library
Artificial intelligence
Evaluate where AI is useful, understand the systems behind it, and make informed choices about models, data, and deployment.
158 articles · Page 11 of 14
Small MoE Models: How Sparse Routing Makes Efficient AI Possible
Small-scale Mixture of Experts: sparse routing lets 47B models match 70B dense equivalents. Mixtral, DeepSeek-MoE, Phi-MoE, efficiency math.
Block Just Cut 4,000 Jobs Because of AI. The Numbers Are Staggering.
Jack Dorsey says AI efficiency is driving Block to slash 40% of its workforce. The real story is what comes next.
How AT&T Cut AI Costs by 90% With Small Language Models
AT&T processes 8B tokens daily on specialized small language models, cutting AI costs ~90% without sacrificing capability. How the architecture works.
Perplexity Computer: A $20B Bet on Model Specialization
Perplexity Computer orchestrates 19 models in one agent. Models are specializing, not consolidating — multi-model orchestration, not commodities.
GGML and llama.cpp Join Hugging Face: What It Means for Local AI
GGML creator Georgi Gerganov joins Hugging Face to secure the future of local AI inference. A landmark open-source move.
Nano Banana 2: What Flash-Speed Image Generation with Web Grounding Actually Means
Google's Nano Banana 2 combines Pro-quality image generation with Flash speed and search grounding. What the architecture implies.
AI Video: From Diffusion to Directors
How AI video generation works: diffusion foundations, temporal modeling, audio sync, and the multimodal architectures behind Seedance 2.0.
OpenAI's Frontier Alliance and GPT-4o Sunset
OpenAI signs multiyear deals with McKinsey, BCG, Accenture, and Capgemini for its Frontier AI agent platform, and retires five older models.
How AI Benchmarks Actually Work
The benchmarks behind AI model claims: SWE-bench, ARC-AGI-2, GPQA Diamond, and more. What they measure, how they work, and what they miss.
Agentic AI Architecture Patterns
A guide to agentic AI patterns: ReAct loops, tool-use protocols, multi-step planning, memory, and multi-agent coordination in production.
Mixture of Experts: Sparse AI Architectures
MoE architectures explained: gating mechanisms, expert routing, load balancing, and why sparse models deliver frontier AI at fraction cost.
Gemini 3.1 Pro Leads 12 of 18 Benchmarks
Gemini 3.1 Pro scores 77.1% on ARC-AGI-2, leads on 12 benchmarks, and doubles reasoning power at no price increase.