Terms Glossary

The cognitive process of selectively concentrating on specific information while filtering out irrelevant stimuli.

Latching onto vivid specifics while ignoring the underlying odds — why a striking case can override what's statistically far more likely.

A systematic, predictable error in how we think — not random noise, but the same wrong turn most minds take in the same situation.

The amount of mental effort required to process information and make decisions.

The mental shortcut of judging quality by how much effort appears to have gone in — useful when you can't see the outcome, misleading when you can.

A mental rule of thumb that trades accuracy for speed — mostly right, fast, and nearly free, which is why the brain leans on them constantly.

Deliberately showing customers the work behind a service — which usually raises how much they value it, sometimes regardless of the result.

The brain's deep difficulty reasoning about chance and uncertainty — a 200,000-year-old instinct that keeps sabotaging modern decisions.

Fast, automatic, intuitive thinking that requires little mental effort.

Slow, deliberate, analytical thinking that requires conscious mental effort.

The cognitive system responsible for temporarily holding and manipulating information during mental tasks.

Systematically measuring whether an AI system's outputs are good — the test suite that tells you if a change helped or quietly broke things.

Reading the brain's electrical activity through the scalp to gauge engagement and response — a direct, if noisy, window into attention.

Turning text, images, or audio into vectors of numbers so that things with similar meaning land close together and can be compared by math.

Measuring exactly where and how long someone looks — turning attention, usually invisible, into data you can actually analyse.

When a model states something fluent, confident, and false — the failure mode that makes ungrounded AI dangerous in serious work.

A model trained on vast text to predict the next token, which at scale turns into fluent drafting, reasoning, and conversation.

Tracing what a model actually did in production — prompts, retrievals, latencies, and costs — so failures are debuggable, not mysterious.

Retrieval-augmented generation that grounds answers in more than text — pulling from video, images, and signals as well as documents.

Building AI that's fair, private, and accountable by design — treating those as engineering requirements, not a compliance afterthought.

Giving a language model the right documents at answer time so it reasons over real, current sources instead of only its trained-in memory.

A simulated stand-in for a real audience, modelled from behavioural data, used to test how people might react before you spend on real ones.

A store built to search by meaning rather than keywords, finding the items whose embeddings sit nearest your query in high-dimensional space.

A shared, data-rich 3D model where every element carries its own information — one coordinated source of truth across every discipline.

Scripting the path from a 3D design to machine instructions, so drawing-to-fabrication that took hours runs in minutes.

Simulating how air or fluid moves around a structure so you can see the loads, the flow, and the failure before anything is built.

How much stronger you build something than it strictly needs to be, to absorb the uncertainty you can't fully model.

The temporary mould that holds concrete until it can hold itself — invisible in the finished building, decisive for whether it gets built at all.

Modelling structures where response stops being proportional to load — where materials yield, geometry shifts, and things behave like the real world.

Tracing a failure back past its symptoms to the thing that actually caused it — then fixing that, not the symptom.

A formwork system that hoists itself up the structure it's building, pour by pour, instead of being dismantled and re-craned at every level.

Letting an algorithm decide where material should and shouldn't sit, given the loads — so the structure earns every gram it keeps.

A range that expresses how precise an estimate is — far more honest than a single number pretending to be exact.

Planning tests so you can vary several factors at once and still know which one moved the result — far more efficient than one-at-a-time.

Testing several changes and their combinations at once to find which mix performs best — A/B testing's higher-dimensional cousin.

A data-driven method for cutting variation and defects out of a process, built on the DMAIC loop: define, measure, analyse, improve, control.

The chance your test actually detects a real effect when one exists — low power means you can miss something true and call it nothing.

Watching a process against control limits so you can tell ordinary noise from a real shift — and only act when something genuinely changed.

Drawing every step a unit of work passes through, so the waiting and waste between the value-adding steps becomes impossible to ignore.

Talking to customers every week as a habit, not a phase — so product decisions trace to fresh evidence instead of last year's assumptions.

The plan for how a product reaches and wins its first users — positioning, channel, and pricing, not just the launch announcement.

The idea that customers don't buy products, they hire them to make progress on a job — so you design for the job, not the demographic.

The smallest thing you can put in front of real users to learn whether the idea holds — built to answer a question, not to impress.

The single measure that best captures the value your product delivers — chosen so that moving it means customers are genuinely better off.

Objectives and key results: a memorable goal paired with a few measurable outcomes that prove you reached it — ambition with a scoreboard.

A map from a desired outcome down through customer opportunities to possible solutions — so you can see what you're betting on, and why.

Letting the product itself drive acquisition, activation, and expansion — the experience does the selling instead of a separate sales motion.

The point where a market clearly pulls a product out of you — demand outpaces your ability to serve it, and you can feel the change.

Whether people keep coming back — the most honest signal a product has, because nothing fakes its way to repeat use.

Scoring ideas by reach, impact, confidence, and effort — a deliberately blunt formula to make trade-offs explicit instead of political.