Skip to main content
← MAIN GALLERYBUILD 043 / 100 · IN THE LAB
BUILD 043
PHASE 05
2026

Design an AI Team

Split complex work across specialized AI agents and people without losing control between them.

WHAT IT IS

A paired public prototype: A is the working tool; B turns the same question into a film, motion argument and interactive web experience.

WHY IT MATTERS

Split complex work across specialized AI agents and people without losing control between them.

RN'S ROLE

RN conceived the question, structured the research and evidence boundaries, designed the information and interaction, implemented the prototype and documented its limits.

WHAT IT CANNOT CLAIM

This independent prototype demonstrates an approach; it is not a deployed client system, professional advice or proof of real-world outcomes.

WORKING PROTOTYPE

You can use and explore this build now. Its public evidence record is still being completed, so treat the experience as a demonstration—not a certified institutional system.

CLAIM CLASSIFICATION INDEPENDENTLY REVIEWED — QUALIFIED SYNTHETIC / DESIGN PROPOSITION

Source: NIST AI Risk Management Framework 1.0. NIST — AI RMF 1.0. Exact locator: GOVERN 1.2–1.7; MANAGE 2.2–2.4. Limit: Context only. People, agents, costs, risks, ownership and failure paths are synthetic; no superiority, productivity or outcome claim is made. This status does not validate real-world effectiveness, safety, compliance, representativeness, cultural authority, professional suitability, or outcomes.

043-A / THE THING

Use it

Use the working tool to explore this question: Split complex work across specialized AI agents and people without losing control between them.

OPEN THE TOOL →
043-B / THE STORY

Watch + explore it

Watch the build-specific film, inspect its motion argument, then use the interactive web version to test: A live multi-agent topology recomputes bottlenecks, supervision, cost and failure paths.

ENTER THE VISUAL BUILD →
THE WORK BEHIND THE ARTIFACT

The build is more than the interface.

Research, planning, evidence, rejected approaches, technical decisions, audits, media, source paths and forward dependencies are retained as part of the build—not left behind in conversation history.

THE QUESTION

Build 043 asks split complex work across specialized ai agents and people without losing control between them.? That question came first and the artifact second, which is the order everything on this page follows. Each section below is either an attempt at the question or a statement of where the attempt stops, and the two are kept visibly apart so you can weigh them separately.

WHICH ROUTE TO TAKE

A and B are two routes into one question rather than a product and its trailer. Open A when you would rather reach your own judgement by operating something and seeing what it returns on your own input. Open B when you would rather understand why the question takes this shape before you touch a control. The two cards below describe each side in its own words, and most people get more out of the pair than either half alone, in whichever order suits them.

WHAT IT LEAVES BEHIND

This is build 043 of one hundred, in phase 05, and it is built to be inherited. What it hands to later builds is concrete rather than thematic: cap:043. That is what makes the sequence compound instead of repeat, and it is why the lineage view is worth reading alongside any single build.

HOW TO CHECK IT

Every claim on this page is meant to be inspectable rather than taken on trust. The public record sets out what changed, what another person can reuse, what supports the system and what the build is not entitled to claim, and the archive retains the research, rejected approaches and technical decisions behind it. A status on this page saying a review is still open is a factual report on the state of that evidence record, and it stays visible until the review closes.

WHERE IT SITS IN THE HUNDRED

Reading one build in isolation gives you the artifact; reading it in sequence gives you the argument. The gallery lets you enter at any number, and the archive holds the syllabus, feasibility audit and release standard that govern the whole programme. If this build interests you, the neighbouring numbers are usually the fastest way to see the reasoning that led into it and the capability it hands forward.

What this build is testing

Split complex work across specialized AI agents and people without losing control between them.

What it leaves behind

cap:043