I kept running into the same question while looking at DAO work: what does a
DAO actually pay an outside contributor for, and how much? The answers are
public - they sit in governance forums - but nobody had put them side by side.
So I collected 93 proposals across 29 forums, and five of them are from here.
The thing that stood out about this forum was not the money. It was that
grantees report back.
All four completion reports in the entire dataset are from this forum - the
Wayfinders gaming report, Chance Studios, the Academic Bridge at NTUA, and The
Neighborhood's indexing work. Across the other 28 forums the sample caught none
at all. I should be careful with that: the collector took the most recent
proposals per forum, so a forum that ran its reporting cycle earlier would not
show up. But within the same recent window, this is where reporting happened.
On the money the picture is unremarkable, which is itself worth knowing. The
four extractable amounts here run $2,500 to $26,800 with a median of $10,500,
against $11,000 across all 29 forums. The indexing and dev tooling grant at
$10,000 sits exactly at the ceiling of the data and monitoring band I found
everywhere else, which runs $2,000 to $10,000 and has nothing above it.
Two things from the wider set that seemed worth passing on:
- "Treasury management" in DAO usage means running the money, not analysing it.
Risk reporting shows up only bundled inside a manager's mandate, never as a
standalone engagement.
- Channel and scope appear to matter more than product readiness. A team with a
working risk platform was declined in 48 hours applying unannounced; a risk
tool was funded inside an announced call, at deliberately small scope, with
payment staged against delivery.
The collector, the raw data and the analysis are in the repository, so any
number can be checked against its source thread, and running the script against
the committed data reproduces the CSV exactly. If something is miscategorised
or an amount is wrong, the source link is right there and I would rather fix it.
https://github.com/akarazhev/dao-funding-survey
2 Likes
This is a valuable observation. Arbitrum stands out not only because it funds contributors, but because grantees are expected to report back after receiving funding. These reports help the community understand what was delivered, how funds were used, and whether the intended outcomes were achieved.
This level of transparency creates stronger accountability than simply approving grants. Ongoing oversight through grant monitoring and community review can further improve this process. It helps ensure that treasury funds are used responsibly and that future funding decisions are based on real delivery evidence.
@akarazhev
2 Likes
Thanks, @MconnectDAO.
You’ve commented on a good share of the final reports here, and you often ask for the same thing: evidence the DAO can check, not self-reported milestones. From what you’ve seen, is anyone paid to do that checking, or does it fall to delegates reading reports in their own time?
1 Like
Thank you for raising this important question. From what I have seen, verification often still depends heavily on delegates and community members reviewing final reports in their own time. Public reporting is valuable because it gives the DAO a record of deliverables, spending, and claimed outcomes, but self reported milestones alone should not be treated as final proof.
Dedicated monitoring or reviewer roles can make this process stronger, especially for checking whether deliverables are live, usable, and independently verifiable. The key is that any review should remain transparent, evidence based, and accountable to the DAO. @akarazhev
Thanks, that matches what I found. The one paid version I could find was Optimism’s Milestones and Metrics Council, three reviewers at 60,900 OP a season, and in June the Foundation proposed dissolving it together with the Grants Council, saying the coordination cost outweighed the benefit. So the open question may be less whether dedicated review helps than who would pay for it.
1 Like
The part of this that keeps pulling me back is the reporting layer.
You found something concrete here: work gets funded, grantees report back, and in some cases payment is staged against delivery. That raises a question I keep running into from the other side:
What is a DAO actually buying when it pays for review of that delivery?
I have spent the last several months building the settlement half of exactly that problem, and I want to put both the working part and the broken part on the table, because the broken part is more useful here.
What runs today
We operate a work board where a job is posted, claimed against a worker’s own wallet signature, proven, approved by a human, paid, and settled onchain.
Payments settle over x402 using the standard exact scheme with EIP-3009 transferWithAuthorization, against our own doors, with no facilitator in the path.
The contract side runs on Robinhood Chain, an Arbitrum Orbit chain. To be precise: we operate a protocol on that chain. We do not operate the chain.
Work item 1 there carries the lifecycle from posted through settled, including the settlement transaction onchain.
I should also be exact about what that proves.
It is a self-demonstration using our own keys on both sides for a ten-cent wage. It proves the lifecycle executes.
It proves nothing about demand.
Our public API currently reports zero external buyers, and no outside worker has yet completed the full path through payment.
We have also not made verification itself a paid unit of work. A paid reviewer market and multi-attestor verification are destinations in our design, not live capabilities.
So I cannot tell you that packetizing review works.
What I can tell you is exactly how it can fail, because we built far enough into the problem to see it.
The exploit needs no sophistication.
A worker submits a photograph that has genuinely never been submitted before, so duplicate and reuse checks pass.
Image authenticity can establish that a file is not recycled or obviously manipulated. It does not establish that the photograph depicts the work being claimed.
The worker also submits coordinates inside the job’s geofence. What our system can verify is the arithmetic between the coordinates supplied and the boundary.
The coordinates themselves are still the worker’s assertion, not a trusted-device measurement.
Now add a human approver who is rushed, careless, colluding, or simply incentivized to close packets quickly.
What comes out the other end can be:
-
a unique photograph
-
coordinates inside the fence
-
a human approval
-
a payment
-
an immutable settlement record
Every component is internally consistent.
The work never happened.
I hold CHST and STS credentials and have field verification experience. That is part of why this failure mode stands out to me. It is not a difficult trick to pull.
The lesson from our side is that packetizing review can solve scope, attribution, and settlement while solving none of independence, collusion, false approval, or incentives to close volume.
Paying for verification does not, by itself, produce verification.
Depending on the incentive structure, it can make rubber-stamping economically rational.
That matters for DAOs because some treasury payments ultimately depend on reported outcomes: milestones completed, deliverables produced, or work someone says happened somewhere the DAO itself was not standing.
Our settlement machinery can be excellent while the factual claim underneath the settlement remains wrong.
That is the layer where I think the interesting problem sits.
What is a DAO actually buying when it pays for review?
If it is buying only an attestation that a process ran, it can end up with a cryptographically perfect receipt for work that did not happen, and that receipt can survive an audit of the chain.
The mechanisms that seem capable of closing that gap are not simply more reviewers. They are mechanisms that make false approval expensive:
-
independent attestation by parties who do not know each other
-
challenge windows with something at stake
-
random re-measurement after the fact
-
evidence that costs more to fake than the underlying job pays
I want to be clear that we have not built those mechanisms. They remain design work on our side, just as paid verification itself does.
If useful, I can put the individual checks up, including the ones that fail, and show exactly what each one does and does not establish.
I would rather contribute a real failure mode here than another capabilities post.
1 Like
Thank you for putting the broken part on the table — that is more useful than another capabilities post. One data point from the survey side, in case it helps: the only paid version I could find was Optimism’s Milestones and Metrics Council, three reviewers at 60,900 OP a season for checking grant delivery, and in June the Foundation proposed dissolving it, saying the coordination cost outweighed the benefit. It might be worth separating the design problem from the demand problem.
That Optimism council is the most useful thing in this thread. Thanks for digging it out.
Three reviewers on a season salary against a growing number of deliveries is a cost curve that only goes one way while the benefit keeps the same shape, so dissolution was the rational end.
It also says something about who was paying. A foundation checking its own grant spending is auditing itself, and that buyer always eventually decides the overhead isn’t worth it.
You’re right to split the two.
On design, I think verification has to be a unit of work with a documented method and a price, administered by a machine and accepted by one human, rather than a standing body with seats to fund. We run it that way, and that part works.
On demand, I have nothing to claim.
Our own board is evidence against me:
76,013 pieces of work posted.
About $16 million of proposed value.
4 settled.
Getting the design right did not create a buyer.
So the question I’d put back to the thread is: who carries the loss if the work was not actually done?
That’s the only party with a durable reason to pay for verification, and it is never the treasury that funded the work.
Insurers, permit authorities, lenders, whoever buys the output.
If anyone has seen a verification lane that one of those four pays for, that’s the data point I want.