cloudparts

AI

Language models, and the agents and apps built on them.

11 concepts · 24 technologies · 1 comparison

Concepts

Start with Large language models
  1. Large language models1 technologyA neural network trained on huge amounts of text to predict the next token; run in a loop, one token at a time, it turns a prompt into answers, summaries or code.
  2. Model hubs1 technologyRegistries for trained models and datasets: each model is a versioned repository of weight files, config and a model card that says what it is, what it was derived from and under which license you may use it.
  3. Running models locally3 technologiesRunning an open-weight model on your own laptop or server: a runtime loads the weights, usually quantized to 4 to 8 bits, into memory and serves an OpenAI-compatible API on localhost, so prompts never leave the machine.
  4. Model routers and gateways3 technologiesA proxy between your app and many model providers: one API and one key in front of OpenAI, Anthropic, Google and others, which routes each request, falls back when a provider fails, and caches, rate-limits, logs and bills the traffic.
  5. AI agents2 technologiesA language model in a loop with tools: it picks a function to call, your code runs it, the result goes back to the model, and the loop repeats until the task is done or a limit stops it.
  6. Agent frameworks5 technologiesLibraries that run the agent loop for you: they call the model, execute tools, keep state between steps, and add checkpoints, streaming, human approval and tracing around it.
  7. Agent tools and integrations8 technologiesHow an agent gets its tools: functions described by a schema, packaged as Model Context Protocol (MCP) servers that any agent can use, found through registries, and authorized per end user so the agent acts with that person’s access.
  8. Retrieval-augmented generation (RAG)7 technologiesRetrieve, then generate: search your own documents for the passages most likely to answer a question, put them in the prompt, and have the model answer from them, citing which passage said what.
  9. Model serving and MLOps5 technologiesThe work between a trained model and a running product: recording every training run and model version, packaging a model with its code and dependencies as a service, and serving it with batching and autoscaling so you always know which model answers and how it was made.
  10. ML frameworks1 technologyLibraries for training and running neural networks: arrays on GPUs and TPUs, automatic gradients, and a compiler that turns Python model code into fast kernels.
  11. JEPA (Joint-Embedding Predictive Architecture)A way to train models by predicting the hidden part of an input as an abstract representation, not as pixels or tokens; Yann LeCun’s proposal for world models that can plan, so far a research line rather than a product.

Technologies

Grouped by the concept each one mainly implements

Commonly confused

LLMs or JEPA?Large language models · JEPA (Joint-Embedding Predictive Architecture)

Worth exploring

All resources →
WebsiteOpenAPI.ToolsPhil Sturgeon and APIs You Won’t Hate, with community contributorsAn open-source directory of more than 300 tools for working with OpenAPI descriptions, sorted into 23 categories such as mock servers (about 25, including Prism, Microcks, WireMock and Specmatic), SDK generators, linters, documentation, gateways, testing and monitoring, and filterable by language and by which OpenAPI versions each supports. Entries arrive by pull request on GitHub; sponsors get priority placement and a badge in their categories, so the order isn’t a ranking.