Google Adds Gemini 3.8 Live and an Extended Thinking Voice Model
Google introduced two real-time voice models: Gemini 3.8 Live for cost-efficient conversation and an Extended Thinking variant that reasons mid-sentence across 97 languages.
5 min read
The archive
Generative AI
New models, what they can do, and how to build with them.
Tokens and Context Windows: What They Mean for Your AI Projects
What tokens and context windows are, how they drive the cost and speed of language model applications, and how to budget context sensibly in real projects.
8 min read
Prompt Engineering Techniques That Hold Up in Practice
Durable prompting techniques for real applications: clear instructions, examples, structure, context, task decomposition, output formats, and testing prompts like code.
8 min read
Fine-Tuning vs. Prompting vs. RAG: Choosing the Right Approach
A practical decision framework for choosing between prompting, retrieval-augmented generation, and fine-tuning, including when to combine them and what each really costs.
8 min read