Cache
Concepts, ideas, mental models, and knowledge snippets. The stuff you want loaded before someone says “just add AI”.
Practical AI, minus the fog machine
Understand LLMs well enough to make better decisions.
theLLMs is a guide site for people trying to use, buy, build with, or explain modern AI systems. It focuses on what the tools do, what they cost, where they fail, and how to test them before trusting them.
Date-scoped sources, explicit methodology, and clear caveats on every serious claim. How this site is made →
Coverage
Broad enough to be useful, weird enough not to become an AI junk drawer
Run
Step-by-step workflows for teams using LLMs in coding, support, search, RAG, evaluation, agents, and internal operations.
Diff
Dated context: what changed, who should care, and what remains unproved., pricing moves, policy changes, benchmark claims, and provider shifts. Diff with a “so what?” attached.
Model and tool comparisons
Decision-first comparisons covering model families, coding agents, inference hosts, vector stacks, eval tools, orchestration layers, and cost trade-offs.
Use cases and business adoption
Where LLMs help, where they fail, what decisions matter, and how teams avoid buying a very expensive autocomplete-shaped theatre set.
Tokens, pricing, evals, and benchmarks
Cost control, quality checks, latency trade-offs, prompt caching, eval harnesses, and benchmark interpretation without cargo-culting leaderboards.
Coding agents and deployment patterns
Hands-on notes for agent loops, tool use, MCP, local development, inference deployment, observability, and production guardrails.
Reader jobs
Questions the site should answer without hand-waving
- What is this thing, really?
- When should a team use it versus avoid it?
- How much does it cost in tokens, latency, maintenance, and operational drag?
- How do you test whether it works instead of trusting a demo?
- What changed, who should care, and what is still unproved?
Browse
Pick the lane closest to the decision you need to make
Cache
Concepts, explainers, comparisons and mental models for the things AI people say too quickly.
Run
Practical workflows: choosing a model, testing RAG, pricing a feature, or running an agent safely.
Diff
Short context pieces that answer what changed, who should care, and what evidence is still missing.
100-page programme
100/100 briefed, drafted, reviewed, and integrated. The full programme is live.
How it is made
The site is transparent about AI-assisted writing, model-labelled writer and editor roles, and human review rather than pretending prose appears by magic.
Search by idea
Find articles by describing what you need
Try "how much do tokens cost?", "run a model on my own hardware", or "stop prompt injection attacks". Search runs in your browser against our article index.
Describe what you are trying to understand.