pressfit.aiGlossary
Research Series Glossary

Stop Prompting and Start Engineering: the Glossary

The AI jargon from Part 1 and Part 2 of our research series, in plain English.

The Basics

Claude
Anthropic's AI model. Every test uses the same model, so the only thing that changes is what surrounds it.
Opus, Sonnet and Haiku
Anthropic's three model tiers, from the most capable and expensive to the fastest and cheapest.
Prompt
The written instructions you give the model.
Agent
A model that decides its own next step. It reads, calls tools and writes, and it keeps going until it thinks the job is done.
Tokens
The units AI usage is measured and billed in, roughly pieces of words. More tokens means a bigger bill.
Deterministic
The same input produces the same steps and the same result every time. Code is deterministic. A model on its own is not.

How Teams Wrap the Model Today

Template
A fixed outline for the deliverable: which sections it has, which fields it needs and how long each part can be.
Skill
A saved set of instructions for one part of a job, which Claude loads when it needs it. Think of it as a playbook page written for the model.
Hook
A check that runs automatically at a set point, for example after Claude writes a section or before a command runs, to make sure the work is right before it moves on. A hook can block the step if something is wrong.
Harness
The program wrapped around the model. It decides what the model can see and what it can touch: which tools it gets, which commands need permission, what runs in a sandbox. Claude Code is a harness. Think of a harness on a horse: it steers and restrains, but the horse still does the walking.
Sandbox
A walled-off space on your computer where some of the agent's commands run, so they can't reach the rest of your files or network.
Deny rule
A hard rule in the harness, such as "never read this folder." It's checked before anything else, and the model can't talk its way past it.
Auto mode
A Claude Code setting where a separate AI checker approves most of the agent's commands, instead of asking a person each time.
MCP (Model Context Protocol)
A standard way to plug a model into tools and data, like a universal adapter. With MCP, Claude can search, read files or call another system. MCP gives the model access. It doesn't decide the order of the work.
HITL (human in the loop)
A person checks the AI's work before it moves on or goes out the door. Teams do it because the model's output isn't reliable enough to ship on its own: someone has to catch what's missing and fix what's wrong. It raises quality, but every deliverable still waits on a person, so you don't get the time savings you were promised.
Headless
Running Claude with no person at the keyboard, on a server or in a container in the cloud.
Session and orchestrator
A session is one running conversation with Claude. An orchestrator is a session whose job is to start other sessions, hand each one a task and check what comes back.
Context window and context engineering
The context window is everything the model can see on a given call: instructions, examples, files and history. Context engineering is deciding exactly what goes into it, step by step.

How We Build It

Business Process Engine (BPE)
Software that runs a business process as a defined series of steps: what happens first, what comes next, what to do when something fails, and when a person needs to approve. The process lives in code, not in a prompt.
LangGraph
An open-source framework for building that kind of workflow as a graph. Each step is a box, the paths between the boxes are written in code, and the graph keeps track of where the job is, so a stopped run can pick up where it left off.
Google ADK
Google's Agent Development Kit, a framework for building agents and coded workflows.
Graph
The workflow itself, drawn as steps and the paths between them. In our posts, "the graph" and "the engine" both mean Pressfit BPE.
Node, step and fan-out
A node, or step, is one box in the graph. A fan-out splits a list into many copies of the same step running side by side, one item each.
Gate
A checkpoint where the run stops for a person to approve before it moves on. Gate 1 is the first one.
Prompt caching
Reusing the fixed start of a prompt across calls, so you pay full price, and wait full time, only for the part that changes.
JSON schema
A strict form the model's answer has to fill in, so code can check every field.

How We Measure It

Run
One full pass of a job, from start to finish.
Same answer twice
We ran the same step twice on the same keywords and counted how often the answer matched.
Agreement
How often one model's answers match another's on the same items. Agreeing with a bigger model isn't the same as being right.
S_det
The share of runs that passed every deterministic code check on the first try.
pass^k
Each job was repeated five times, and pass^k is the chance that all k of those repeats pass (tested from k = 1 to k = 5). Passing four out of four (k = 4) is our main bar.

The Names in Our Charts

Claude alone (A0)
Claude gets the brief and the source files, and nothing else. It decides how to do the job, step by step. This is how most teams use AI today.
Claude + template
Claude alone, plus a template that sets the structure of the deliverable.
Claude + skills and hooks (A-Full)
The template, plus our skills for each part of the job, plus hooks that check Claude's output against the template before it moves on. This is Claude with everything turned on.
Pressfit BPE (G and G1)
Our Business Process Engine: our own code execution, running on LangGraph, with MCP connecting it to tools and data. Code runs the steps in a fixed order and keeps the state, while still pulling the data each job needs and reusing what earlier jobs gathered. Claude works only at the steps that need judgment or writing, and every result has to pass the Red Dev Test team's tests before it is delivered. G and G1 are two versions of the engine.