Library
Explainers, decks, playbooks, and tools. Everything on the site, in one place.
Browse resources
15 resources
An LLM is not a database.
FYILLMA large language model generates text using patterns learned during training and the information supplied with a request. A product built around it may add search, memory, and tools.
An agent can choose what happens next.
FYIAgentAn AI agent is a system that uses a model to choose steps and tools while working toward a goal. Results from one step can change what it does next, within the permissions and limits of the surrounding software.
The software around the model matters.
FYIHarnessA harness is the software that coordinates calls to a model and connects them to tools and results. Its scope varies: context handling, checks, logging, and recovery may sit in the harness or in services around it.
Context is the information supplied for this response.
FYIContextContext is the information available to a model when it generates a response, including instructions, messages, and selected tool results or documents. A product can carry information across turns, but stored history is not necessarily included in every model call.
RAG helps AI answer from a source.
FYIRAGRetrieval-augmented generation finds relevant information and supplies it to a model before an answer is generated. It can ground an answer in selected sources, but retrieval mistakes and unsupported claims are still possible.
A confident answer can still be wrong.
FYIHallucinationsA hallucination is generated content that is false or unsupported but presented as if it were established. Fluent wording, a confident tone, or a detailed explanation does not by itself verify the claim.
AI 101: Beyond the Chat Box
Deck8 slidesA plain-English introduction to what AI can do, where it breaks, and how to find a useful first move.
Where AI Belongs
Deck8 slidesA practical way to choose whether AI should assist, recommend, automate, or stay out of a product decision.
Earned Autonomy
Deck8 slidesA leadership briefing on expanding AI responsibility only as evidence, safeguards, and recovery improve.
The AI evaluation plan product teams can actually run
PlaybookWorksheet + PDFReplace a vague quality score with a product-specific system for deciding whether to launch, expand, redesign, or stop.
The earned-autonomy ladder for AI products
PlaybookWorksheet + PDFDecide what an AI system may observe, recommend, draft, or execute—and what evidence it must earn before receiving more authority.
The one-page AI opportunity brief
PlaybookWorksheet + PDFTurn a vague AI idea into a testable product bet with a named decision, observable value, and an explicit stop rule.
Decision Compiler
ToolRuns in your browserTurn an ambiguous AI idea into the right product role, a first test, useful measures, and clear safeguards.
HTML presentations
SkillCodex + ClaudeTurn a rough idea into a self-contained, responsive presentation with strong narrative, presenter notes, and clean PDF export.
MiniMax H3 Max
Sourcefal.aiIn its September 8, 2026 benchmark, fal reports 15.17 seconds of inference for MiniMax H3 Max and 8.44 seconds for H3 Max Turbo on a 15-second, 768p text-to-video request. Vendor-reported benchmark, not an independent test. Queueing, prompt preprocessing, and video encoding add to the caller's wait time.