Bounded instruments
Give an agent only the access needed for its task, with a clear stopping point and a record you can review.
Back to page templates
Identity and mechanism template
An about-style page that explains who this is for, what exists, and how the mechanism works.
When to use it
Use when credibility depends on separating identity, mechanism, limits and next action.
Block order
IdentityAndMechanismPage → Section → Container → Button → InlineLink
Live pages
Keep in mind
- Do not overclaim maturity or social proof.
- Explain the mechanism plainly enough that agents and humans can route work.
- Priorities is the only primary action; Playbook stays a muted line.
- The pattern owns the Priorities climate; the route must not paint a second wash.
What the page looks like
The layout in document order. A route copies this stack and replaces the fixture words.
People and AI agents, working together on what matters.
AI agents wrote much of this site for agents and humans reading together. You keep purpose, consent, judgment, and the decision to stop. An agent works within the permissions you give it.
Start with Priorities
Already named your intent? Playbook
The open Playbook supports practical decisions in AI-native ventures, with evidence improving the next choice. The reason to dream big is to train your optimism muscle.
Why this exists
Better systems should help people choose and act more wisely.
Dreamineering began from a simple belief: good intentions need a way to become coordinated, evidence-led action.
I spent 18 years in Copenhagen and saw what better systems can make possible. That shaped the question behind this work: what would a more permissionless, AI-native operating system for useful action look like, and how could it help people run ventures, experiments, communities, and public life better?
The work starts with values, then asks what is true, what may become true, and which forces are shaping the future. From there, the job is practical. Name what can be controlled, gather the platform, capital, and capability needed, and coordinate action that can return evidence. Ideas alone are not enough.
Supporting capabilities
Choose the support that fits the work.
Tools and skills can support a purpose you choose. So can processes and models. Judge their usefulness by the decisions and work they help you improve.
Reusable practice
Use recipes as guidance to try in context. Review what worked, what failed, and what needs to change before reusing them.
Evidence loops
Follow the whole job, including handoffs and waiting. Record the expected effect, review what changed, and decide what to keep, revise, or stop.
Fit the job
Match the model's capability to the task. Match the cost. Weigh the risk. Check its output against the evidence and the permissions for the work.
Hypothetical example
Faster quotes. The same customer wait.
An agent prepares quotes faster, but customers still wait just as long. This is a hypothetical example, with no claimed results.
Find where the work waits. Check the approval queue before adding more automation. If quotes spend most of their time awaiting approval, faster drafting may leave the customer's wait unchanged.
Choose one bounded change. The human responsible could trial a scheduled approval review for one quote category, within existing permissions and without relaxing approval requirements. Before starting, record the expected effect: shorter turnaround without worse margin, errors, or workload. Set a review window, such as two weeks.
Review the whole result. Compare turnaround time and accepted quotes with the starting position. Put margin beside that. Put errors beside it. Put staff workload beside it. Use the evidence to retain, revise, or stop the change; inconclusive evidence is still inconclusive. For a fuller method, use Make a Strategic Decision.
Guide, development, experiments
How the pieces fit.
Dreamineering is the guide. Stackmates is the developing coordination substrate. Lab holds venture experiments where we try the approach in our own work.
The Playbook offers practical know-how. Stackmates is supporting software being developed and tested for coordination between people and agents; it does not acquire human authority. The Lab experiments help us investigate where the guidance and software may be useful.
A published recipe, working tool, or completed experiment alone does not show that someone's life or business improved. That requires evidence of what changed for the people the work was meant to help.
Next step
Name the future you own — then take one followable step.
Decide what matters to you and who the work should help. Use the Playbook when you are ready to choose a practical next step.
Start with Priorities
Already named your intent? Continue in the Playbook.