For the People Who Run Grant Rounds
Most rounds can only tell, not show. Ask why one project won and another lost, and the answer is a story, not a record anyone can check.
CROSS, GRAIN, and WALKRI are a small set of free, open standards that make a round's decisions hold up: comparable, checkable, and yours to defend.
The Label
"Is this project open source?"
A yes you have to take on faith.
The Instrument
"Does the primary codebase carry an in the root of its public repository?"
Qualifying: MIT, Apache-2.0, GPL. Non-qualifying: available on request, or open-source promised later. Evidence: a link to the LICENSE file. A yes anyone can verify.
The same question. The difference is whether it measures anything.
It does not have to be either-or. You were never given a standardized way to choose which fields are stories and which are instruments. Now you choose, field by field: a story where you want one, an instrument where you need one.
Why can a round only tell, not show? The data it runs on was never set up to be checked. It is the old , with one wrinkle: the risk is not only at the front door. Every step of a round is a place the data can lose its grip, and a place you can now set deliberately. This is not a knock on anyone's judgment; it is simply what happens when the structure was never there, which is why it looks the same in a foundation, a development bank, or an on-chain round.
Follow One Round Through Its Life.
At each step the same choice opens up, narrative or instrument, and each standard is what lets you make it on purpose instead of by default.
Step 1 · The Operator Sets the Round Up
By Default
The round takes shape as applications arrive. Criteria settle by feel and shift mid-round.
By Design
The round commits, before it opens, to what it funds, how it will judge, and which level of change it is paying for.
Decided once, in the open, instead of improvised under pressure. CROSS commits the round's obligations and gates before it opens. GRAIN declares what the round is buying, including the level of change in its theory of change. WALKRI makes each committed field an instrument.
Step 2 · The Grantee Applies
By Default
You apply into fog, guessing what counts, writing to impress.
By Design
You see the exact measures before you start, and write to meet them.
The target is declared, not inferred. WALKRI is why the measures you see are real instruments, not vague asks. GRAIN is how you state your pathway, the causal claim from your work to the change.
Step 3 · The Reviewer Evaluates
Judged by Impression
"Did they deliver what they promised?"
A response. Two reviewers can read it two ways and never know they disagreed.
Judged From an Instrument
"Are the proposal's milestones marked complete, each with a verifiable artifact?"
Qualifying: a release tag, a published report, an on-chain transaction. Non-qualifying: "on track", "nearly done", a screenshot. Evidence: a link to each artifact. A reading two reviewers reach the same way.
The same question, asked so it measures something. This is the difference the whole page turns on. WALKRI turns the question into an instrument. CROSS holds you to the gate the round committed to, not a moving one.
Step 4 · The Analyst Reads What Came Out
By Default
Outcomes arrive as a pile of numbers with no theory behind them. You cannot tell whether what the program predicted came true, and results at different levels, an output here, a long-term goal there, get compared as if they were the same thing.
By Design
Every grant declared its pathway first: the causal claim from the work to the change, the assumptions it rests on, and the level of outcome it aimed at. So predicted lines up against actual, across rounds and across cycles, and you can see which theories held and which assumptions broke.
So each cycle tests the last one's model, and the model gets sharper instead of starting from scratch. GRAIN carries the shared format and the theory of change, so results line up and can be tested against what was predicted. CROSS makes each round's record a structured attestation, not a scattered report.
is built into all four standards by design, not bolted on. See how each one carries it →
Step 5 · The Institutional Funder Audits
By Default
To check a program you have to ask, then trust what you are handed.
By Design
Obligations and a record, published and checkable, on the record without anyone requesting it.
Checkable by default, not on trust. CROSS lets you audit against the published obligations. CRAFT is why it is checkable at all: every stage has to be readable from the published specification.
Step 6 · The Platform Provider Builds On It
By Default
Every integration is bespoke. Adopting one platform locks you to it.
By Design
A public-domain common layer any tool can read and build on.
CC0, no licensing, no lock-in. CROSS, GRAIN, and WALKRI are one CC0 layer you can build on. CRAFT is the meta-standard the others are built to satisfy.
All of this lives in front of whatever collects the data and behind it, never in the collection itself. A form today, an agent reading documents tomorrow: the standards do not care which. They set what has to be captured before anything is collected, and hold the result to a standard after, so what comes out is something a machine, and an AI, can read by construction. The round you run now is ready for the tools that are coming.
That is the shape of it. The working model underneath, what you choose to collect, the round lifecycle, and the evidence it rests on, is one step deeper.
How a Round Works →Tell us what you run, and we will help you set it up.
Or go straight to the standards: the standards assistant (bring your own key), or read CROSS, GRAIN, and WALKRI.
Open Standards
CROSS, GRAIN, and WALKRI are dedicated to the public domain. Read them, use them, build on them. GRAIN’s specification is in preparation.