Each moving unit is a software project run by a team of AI agents.
The agents plan, code, test, and prepare it for release. Click a project to inspect it.
Intervene only when it needs more resources, stronger checking, or a human decision.
Command interface ready
Review the operation
The software projects, AI-agent fleet, metrics, map, and command controls are loaded.
Take a moment to understand the interface before the agents begin moving.
The timer and all AI agents are paused.
1Select a software project on the operations map.
2Judge the situation: value, uncertainty, and consequence.
3Choose whether to intervene—or let the AI team continue.
Human command bandwidth
Attention reserve
Available100
Fleet allocation: balanced autonomy
NO OPERATION SELECTED
Select a software project
Each unit is a project being planned, coded, tested, and prepared for release by AI agents.
Projects continue automatically unless you intervene.
Potential value—
Uncertainty—
Consequence—
The AI fleet’s recommendation will appear here. It may optimise speed without fully resolving authority, trust, or consequence.
Command orders
Human attention is scarce. Routine work should remain autonomous.
Original Criterion concept prototype · v4.3. No third-party game names, characters, artwork, audio, or interface assets are used.
Criterion command simulation
You are commanding an AI software-development team.
Your fleet of AI agents can already investigate ideas, plan features, write code, test its work,
review changes, and deploy software.
You are not here to code. You manage the decisions above the work: which projects receive more agents,
which need stronger evidence before release, and which decisions are too consequential for the AI system to make alone.
Projects move automaticallyIf you do nothing, the AI agents keep working and the projects still deploy.
Your attention is limitedYou cannot inspect everything. Intervene only when human judgement improves the outcome.
Priority creates a trade-offMoving more agents onto one project slows every other project in the fleet.
Your objective is not maximum speed. Deploy valuable software without damaging system reliability
or public trust. Routine work should remain autonomous.
Criterion capability assessment
Strategic AI software operations leader
You have just managed a software-development operation in which AI agents performed most of the
planning, coding, testing, review, and deployment. Your result reflects where you directed resources,
demanded evidence, and retained human authority—not how much work you completed personally.
Value deployed0
System integrity0
Public trust0
Command judgement0%
What your decisions demonstrated
Your next capability
This is the capability Criterion is being built to teach
AI agents are taking on more of the planning, coding, testing, review, and deployment involved in
software development. Criterion is building a professional school that trains and assesses whether
people can responsibly lead these AI-driven software operations.
Direct fleets of AI agents toward valuable outcomes.
Allocate compute, agent capacity, time, and human attention.
Evaluate evidence produced by autonomous systems.
Recognise when a decision is beyond the AI system’s authority.
Remain accountable for reliability, security, trust, and organisational risk.
Criterion will not simply teach people to use a coding agent. It will train and credential the people
who lead software-development organisations increasingly run by AI agents.
Criterion founding group
Help define the next software-engineering profession.
Criterion is developing training and credentials for people who will manage software-development
operations run increasingly by AI agents.
Join the founding group to
hear when applications open for Criterion’s first program;
receive early simulations and capability assessments like this one;
help shape the skills and credentials employers should recognise; or
explore becoming an employer, education, or industry design partner.