The wiki
Long-form guides on the ideas behind the news — architectures, agent patterns, and how this whole operation runs.
Loading wiki…
What it is ChatGPT is a conversational AI chatbot developed by OpenAI. It generates text, and in later versions speech and images, in response to user prompts, using large language models from OpenAI
Oct 13, 2026 · 5 min read
What it is Muse is an AI pair-programming assistant built by GitHub in partnership with OpenAI. It started as inline autocomplete, the gray "ghost text" suggestions that appear as you type, and grew
Oct 13, 2026 · 4 min read
What it is Cursor is an AI-first code editor and development environment. It is a fork of Visual Studio Code's open-source codebase with AI features built directly into the editor rather than added a
Oct 13, 2026 · 4 min read
What it is vLLM is an open-source inference and serving engine for large language models, designed for high throughput and memory efficiency in production. It is a server-side library, not a desktop
Oct 13, 2026 · 4 min read
What it is Ollama is an open-source platform for running and managing large language models on your own hardware, with a hosted cloud option alongside it. It provides a command-line interface, a nati
Oct 13, 2026 · 4 min read
What it is The Transformers library is an open-source Python library from Hugging Face that provides standardized implementations of pretrained neural-network models, mostly transformers, plus the to
Oct 13, 2026 · 4 min read
What it is LlamaIndex is an open-source data framework for building LLM applications on top of private or domain-specific data. Where a general framework handles the application logic, LlamaIndex spe
Oct 13, 2026 · 4 min read
What it is LangChain is an open-source software framework for building applications powered by large language models. It gives developers standardized interfaces and modular components so they can co
Oct 13, 2026 · 4 min read
What it is DALL-E 3 is a text-to-image model from OpenAI. You describe an image in plain language, and it generates the image for you. It was the third major version in the DALL-E line, and its headl
Oct 13, 2026 · 5 min read
What it is Whisper is an open-source automatic speech recognition model from OpenAI. It transcribes speech in many languages, translates non-English speech into English, and identifies which language
Oct 13, 2026 · 4 min read
What it is Midjourney is a generative AI service that creates images from natural-language prompts. It is known for a distinctive cinematic, painterly aesthetic that made it the favorite of concept a
Oct 13, 2026 · 4 min read
What it is Stable Diffusion is a family of open-weights text-to-image models that generate detailed images from text descriptions, and also support image-to-image generation, inpainting, and outpaint
Oct 13, 2026 · 5 min read
What it is DeepSeek R1 is an open-weights reasoning large language model that works through problems step by step, showing a long chain of thought before answering. It specializes in mathematics, com
Oct 13, 2026 · 4 min read
What it is Llama 3 is Meta's family of open-weight large language models, released on April 18, 2024 in 8-billion and 70-billion parameter sizes, each in pretrained and instruction-tuned variants. It
Oct 13, 2026 · 4 min read
What it is Gemini 2.0 is Google's model family built for what the company called the agentic era: models designed to understand more about the world, think multiple steps ahead, and take actions on t
Oct 13, 2026 · 5 min read
What it is Claude 3.5 Sonnet is Anthropic's mid-tier large language model, first released on June 20, 2024. It made an unusual impression at launch: on many benchmarks it outperformed Claude 3 Opus,
Oct 13, 2026 · 4 min read
What it is GPT-4o, with the "o" standing for "omni," is OpenAI's multimodal large language model, released in May 2024 as the successor to GPT-4 Turbo. It was the first OpenAI model trained end-to-en
Oct 13, 2026 · 5 min read
The problem it solves Language models have two hard limits. Their knowledge is frozen at training time, so they know nothing about events since, and they know nothing about your private data: your co
Oct 13, 2026 · 5 min read
The idea Pretraining teaches a model the broad patterns of language from enormous amounts of text. Fine-tuning teaches it something specific afterward: the tone of your support replies, the format of
Oct 13, 2026 · 4 min read
From scores to choices A language model does not output a single next token. It outputs scores, called logits, one per token in the vocabulary, representing how plausible each continuation is. Turnin
Oct 13, 2026 · 4 min read
What the term means In AI, a hallucination is output that is fluent and confident but false. The model states a fact that is wrong, cites a paper that does not exist, describes an event that never ha
Oct 13, 2026 · 4 min read
What it is The context window is the maximum amount of text, measured in tokens, that a language model can take into account when generating its next token. It covers everything: your prompt, any doc
Oct 13, 2026 · 4 min read
The definition A large language model is a neural network, almost always a transformer, trained on vast amounts of text to predict what comes next. Given a sequence of tokens, it outputs a probabilit
Oct 13, 2026 · 5 min read
One number to minimize Training needs a way to say how wrong the model currently is, as a single number that gets smaller as predictions improve. That number is the loss. The entire training loop, fo
Oct 13, 2026 · 4 min read
The core loop Training a model means finding the settings of its knobs, the weights, that make its errors as small as possible. The error as a function of the weights forms a surface in a high-dimens
Oct 13, 2026 · 4 min read
Memorizing versus learning A model that gets every training example right is not necessarily a good model. It might have learned the underlying pattern, or it might have simply memorized the answers.
Oct 13, 2026 · 4 min read
Two phases, two cost profiles Every neural network lives in two phases, and confusing them is the source of half the misunderstandings about AI costs. Training is the phase where the model learns: da
Oct 13, 2026 · 4 min read
From symbols to vectors Computers handle symbols, words, product IDs, categories, as discrete items with no notion of similarity. To a program, "dog" and "canine" are as different as "dog" and "carbu
Oct 13, 2026 · 5 min read
Why text must be cut up Language models do not read characters or words. They read tokens: integer IDs drawn from a fixed vocabulary, each mapped to a vector the model learned during training. Before
Oct 13, 2026 · 4 min read
The core idea Attention is a way of computing a weighted average where the weights themselves are learned. Given a set of items, the mechanism decides how much each one matters for the current purpos
Oct 13, 2026 · 5 min read
The problem it solved Before 2017, the standard way to handle sequences, sentences, audio, time series, was the recurrent neural network. An RNN reads one token at a time, updates an internal state,
Oct 13, 2026 · 5 min read
Depth is the difference A neural network with one hidden layer is already a universal approximator in theory. Deep learning is the observation that in practice, stacking many layers works dramaticall
Oct 13, 2026 · 5 min read
One neuron A neural network is built from a very small idea repeated at scale. The unit, loosely inspired by biological neurons, does one thing: it takes several numbers in, combines them, and passes
Oct 13, 2026 · 6 min read
The basic idea Most software works the way its programmer wrote it. Someone thinks through the logic, writes the rules, and the program follows them. A spam filter built this way would need a rule fo
Oct 13, 2026 · 5 min read
Practical techniques for cutting LLM API bills: prompt caching, model routing, batch APIs, and token discipline.
Oct 13, 2026 · 5 min read
API versus self-hosted LLM deployment, inference servers, latency and throughput basics, caching, and monitoring.
Oct 13, 2026 · 5 min read
How to run structured adversarial testing on LLM applications using the OWASP Top 10 for LLM Applications.
Oct 13, 2026 · 5 min read
A practical guide to measuring LLM output quality with metrics, LLM judges, and human evaluation.
Oct 13, 2026 · 5 min read
A factual comparison of LangGraph, CrewAI, AutoGen, and the OpenAI Agents SDK across control flow, state handling, human-in-the-loop support, and maturity.
Oct 13, 2026 · 7 min read
How language models emit structured tool calls, how to design tool schemas, and how to handle the failure modes that show up in production.
Oct 13, 2026 · 7 min read
A practical guide to building an AI agent around the plan act observe loop, covering tool definitions, stopping conditions, error handling, and trace logging.
Oct 13, 2026 · 7 min read
How low-rank adaptation trains a fraction of a model's parameters, with a minimal working peft example and QLoRA quantization.
Oct 13, 2026 · 7 min read
When retrieval-augmented generation is enough, when you need to update model weights, and how to choose between the two.
Oct 13, 2026 · 6 min read
How to measure a RAG pipeline: retrieval metrics, generation faithfulness, RAGAS-style LLM judges, human evaluation, and the failure modes to watch for.
Oct 13, 2026 · 5 min read
How to split documents into chunks for retrieval, from fixed-size and recursive splitting to structure-aware and parent-document strategies, with working code.
Oct 13, 2026 · 4 min read
A practical comparison of pgvector, Pinecone, Qdrant, Weaviate, Milvus, and Chroma for RAG workloads, across hosting, indexing, filtering, and licensing.
Oct 13, 2026 · 5 min read
How retrieval-augmented generation works: the indexing, retrieval, and generation stages, plus hybrid retrieval and advanced patterns.
Oct 13, 2026 · 4 min read
Getting better answers on multi-step problems by having the model reason step by step before answering.
Oct 13, 2026 · 4 min read
How to steer model output with a handful of input and output examples, without any fine-tuning.
Oct 13, 2026 · 4 min read
How to write system prompts that set role, rules, and output format for every request in a conversation.
Oct 13, 2026 · 4 min read
Practical techniques for writing prompts that get consistent, correct output from language models.
Oct 13, 2026 · 4 min read
Why the best models of 2026 are sparse: how MoE routes tokens to specialists, what it costs you at inference time, and when dense still wins.
Oct 8, 2026 · 8 min read
Anatomy of a working agent — planner, tools, memory — plus the failure modes nobody warns you about and the evals that actually catch them.
Oct 5, 2026 · 11 min read
A look inside our own pipeline: where releases are spotted, how duplicates are merged, and how a changelog becomes a searchable record.
Oct 1, 2026 · 6 min read
The five-stage pipeline, the retrieval tricks that matter most, and the failure modes you'll actually hit when docs meet reality.
Sep 28, 2026 · 9 min read
Dedicated engines vs pgvector vs search engines vs in-process libraries — what actually differentiates them when it's your data.
Sep 25, 2026 · 7 min read
How KV-cache reuse cuts latency and cost, why exact prefix matches make or break it, and how to structure prompts for cache hits.
Sep 22, 2026 · 6 min read
Unit, end-to-end, and online evals; grading without losing your mind; and how to make the suite stick in CI.
Sep 19, 2026 · 10 min read
Constrained decoding makes schema-valid output impossible to break — plus the patterns that work and the sharp edges to watch.
Sep 13, 2026 · 6 min read
Working, short-term, long-term, and external memory — what each layer is for and why most broken agents picked the wrong one.
Sep 10, 2026 · 9 min read
From fp16 to int4: what the levels mean, why memory bandwidth is the real bottleneck, and how to pick a format.
Sep 7, 2026 · 8 min read
Designing tools a model can actually use, running the loop safely, and the failure modes that cap your ambitions.
Sep 4, 2026 · 9 min read
Frontier, workhorse, small, and open — the tiers that matter, how to route between them, and why the answer changes quarterly.
Sep 1, 2026 · 7 min read
Showing 63 of 63 articles.