Tag
ai
Published articles tagged ai.
Vector search and embeddings in MongoDB
Semantic search stores embedding vectors next to documents and finds nearest neighbors with $vectorSearch. This chapter generates embeddings through Voyage AI, defines vector indexes on Atlas, runs hybrid queries with $rankFusion, and sketches a small RAG retrieval function in TypeScript.
A short history of Artificial Intelligence
Artificial intelligence did not start with chatbots. It went through decades of bold predictions, two funding collapses, and a slow shift from hand-written rules to learning from data. This overview follows that path from the 1940s to the language models of the late 2010s.
Agentic AI
A chat assistant answers. An agent acts: it plans, calls tools, reads the results, and keeps going until a goal is reached. This overview explains the agent loop, tools, and the harness around the model, how agents evolved since 2022, and where workflow tools like n8n and personal agents like OpenClaw fit.
AI and data privacy
Every prompt sends data somewhere. This overview follows that data: where it goes, how long it is kept, whether it trains future models, and what data protection law requires. It ends with practical rules for people and companies that want to use AI without leaking personal or confidential information.
AI basic concepts
Model, training, token, prompt, context window, hallucination: a small set of terms explains most of what is said about AI. This overview defines each one in plain language and shows how they fit together, without math and without a specific product in mind.
AI glossary
A reference of the terms that come up when people talk about AI, from token and context window to harness, RAG, and alignment. Each entry gives a short definition and points to the article in this collection that explains the idea in more depth.
AI in software development
Software development changed faster with AI than almost any other kind of work. This overview follows the path from code completion to chat in the editor and autonomous coding agents, and covers what the tools do well, where they fail, and the habits that keep AI-written code trustworthy.
AI regulation
Governments regulate AI in very different ways. The EU has a single law that sorts AI by risk, the United States relies on executive orders and state laws, and China regulates specific services. This overview explains the main approaches as of late 2026 and what they mean for companies that build or use AI.
AI safety and risks
The risks of AI range from a chatbot inventing a fact to fears about systems that people can no longer control. This overview sorts the risks into groups, explains each in plain terms, and describes what companies, governments, and users do about them, including the open debate about human-level AI.
AI versus non-AI
A traditional program follows rules a developer wrote. An AI system follows rules it learned from examples, or instructions written in plain language. This comparison shows what that changes for correctness, testing, cost, and failure, and when each approach is the right one.
Evaluating AI models
Every new model claims top scores. This overview explains what benchmarks and leaderboards actually measure, why their numbers need care, and how to build a small evaluation of your own that answers the question that matters: is this model good enough for my task?
Frontier AI companies
A handful of companies build the most capable AI models, and a few more supply the chips and data centers those models run on. This overview sorts the leading companies by the role they play, names their main products, and explains how the layers depend on each other.
Generative AI beyond text
The same advances that produced chat assistants also generate images, speech, music, and video. This overview explains how each kind of media is generated, what the tools are used for, and the questions they raise about copyright, consent, and telling real content from generated content.
How large language models are built
A chat assistant is the end of a long pipeline: collecting text, training a model to predict the next token, teaching it to follow instructions, aligning it with human preferences, and testing it before release. This article walks through each stage and what it adds to the finished model.
Open-weight versus closed models
You can rent a model through a provider's API or download open weights and run the model yourself. This comparison explains what "open" really means for AI models, how the two options differ in capability, cost, privacy, and control, and when each one is the better choice.
Prompting fundamentals
The same model can give a vague answer or an excellent one, depending on what it is told. This overview covers what makes a prompt work: a clear task, enough context, examples, a defined output, and a habit of testing and revising, plus why context matters more as models get stronger.
Retrieval-augmented generation explained
A language model only knows its training data. Retrieval-augmented generation looks up relevant passages in your own documents and hands them to the model with the question. This article explains the indexing and query steps, where RAG systems go wrong, and how RAG compares with fine-tuning and long context windows.
The cost of AI
AI has a price at every level: per token for a developer, per month for a user, billions for a training run, and terawatt-hours for the power grid. This overview explains where the money and energy go, why prices per task keep falling while total spending rises, and how to keep your own costs under control.
Foundation Model Providers (LLMs & APIs): Companies, Models, and Use Cases
Using a language model in your own software means picking a provider, a model family, and a way to call it. This page explains the ways to access models, how an API call and its pricing work, which providers offer which model families, and how to match a model to a use case.