There is no shortage of "what is an LLM" explainers, and we are not going to write another one. This section covers the parts of AI infrastructure that change a real software decision — explained so a beginner can follow it and a builder still learns something. Two rules: we only write about things we run ourselves or can tie to the 300 tools we research, and nothing gets a number unless we can show you where it came from.
What the Model Context Protocol actually does, drawn out rather than described — plus the 35 tools we partner with that ship one, checked against their own docs, plan and price included.
Credits, caps, add-ons and tiers — where the AI in your software is really billed. Pulled from our own pricing research on 300 tools and quoted verbatim, including the vendor whose pricing page never mentions AI at all.
What the file is, what we put in ours across six sites, and the honest answer on whether it does anything — from someone who publishes one rather than theorises about it.
We are not observers of this stuff. We publish an MCP server, an open dataset and a structured answer file that AI assistants read, and we maintain hand-researched dossiers on 300 business tools. That machinery is what these explainers are built on:
See them: our MCP server · the Atlas · open dataset · how we review.
This section grows slowly and on purpose. A topic earns a page here when we can say something first-hand about it or connect it to partners we already research: agent-to-agent standards, what "AI-native" actually means on a pricing page, how AI features change what software costs, and how assistants decide which sources to cite. What will never appear here: recycled model explainers, prediction pieces, and tool round-ups we haven't researched.
If there's an AI topic you keep getting vague answers about, tell us — that's how this list gets written.
New dossiers, cost-traps we found, and what actually changed in AI tooling — no hype, no sponsored-disguised-as-advice. Unsubscribe anytime.