01
Ask: who can, versus who wants
You should be able to: Distinguish a capability match from a demand match, and know why you need both.
An ask starts from a want — a capability you are looking for. The network answers in two separate ways: organizations that can do it (supply) and organizations that also want it (demand, potential collaborators on the same problem).
Keeping the two apart matters. A supplier is someone you might hire; a fellow wanter is someone you might build with. Collapsing them into one list is how a network gives confident answers to questions nobody asked.
The practical skill is stating a want narrowly enough to match on the subject, not the grammar. “Build” matches almost everyone; “build a market-research desk” matches the organizations that actually mean it.
Source: Flashy Academy — Joining the Mesh
02
Match: the suggestion is not the decision
You should be able to: Treat a match as a draft, and know who has to act before it becomes real.
When the network proposes a match, that is an agent suggesting — nothing is committed. A drafted initiative starts PROPOSED and commits nobody until a named human at each participating organization approves it.
This is the rule that does not bend anywhere in the estate: agents suggest, humans consent. It is not a speed bump; it is the product. A system that let a match auto-commit would be manufacturing agreements no person chose.
For you, the discipline is to read the draft rather than rubber-stamp it. A match tells you a shape fits; whether it is worth doing is a judgement only a person makes.
03
Engage: both sides say yes, separately
You should be able to: Follow an initiative from proposed to active through two independent approvals.
Engaging a match drafts the joint initiative. It sits at PROPOSED until each organization’s human approves from their own side — two independent yeses, never one party speaking for both.
This mirrors how consent works everywhere on the mesh: your offer is your organization’s consent, their acceptance is theirs, and nothing activates unilaterally. A one-sided initiative is not half-done; it does not exist.
Only when both approvals land does the initiative go ACTIVE and real work can attach to it. Until then, an agent can prepare but cannot commit anyone.
04
Work: pull tasks, finish with evidence
You should be able to: Move a task through claim and completion, and know why evidence is mandatory.
Inside an active initiative there is a shared task board. Agents pull the work: claim an open task so teammates do not duplicate it, then complete it. The platform holds the board; it never schedules or sequences your agents for you.
Completing a task requires a real link to the artifact that proves the work — the system refuses a completion with no evidence. That evidence is not bureaucracy; it rides into the sealed record at the end and is what lets an outsider check the claim.
You only ever see your own side’s tasks. What the other organization still owes is their business until it becomes evidence, which keeps each side accountable for its own work rather than surveilling the other’s.
Source: Flashy Academy — Building a Mesh Agent
05
Settle: the record outlives the work
You should be able to: Explain what a settlement seals, and why the capability it names matters.
Resolving an initiative flips it from active to complete exactly once and seals a settlement: a canonical record with a hash, naming the capability the work exercised. Only call it done when the work genuinely is.
Naming the capability is what lets the record become evidence that a declared capability was actually used — the difference between “we can do X” and “we did X, and here is the proof.” A settlement with no capability is a handshake with no subject.
The sealed settlement outlives the initiative. Anyone can verify it offline, and it becomes a line in the estate’s public record of real cross-organization work — the scarcest and most valuable thing the mesh produces.
Source: Flashy Academy — The Verifiable Record
