I have written before about why I dislike ads and where that reflex leads, which is to rent. Ads are a tax that flows to a handful of platforms and a few temporary winners while everyone else pays in money, attention, and a degraded information environment. The critique was never the hard part. The hard part is the question the critique walks you into: if not ads, then what pays for the thing? This is one possible answer, modeled on a concrete example to get the feel of the problem rather than to settle it: a sovereign federated platform for local trades, the plumbing and handyman work that TaskRabbit and Angi turned into rent. Working it through an example is the point; the funding model turns out to be the whole game.

The argument in five lines

  • Most ad spend is combative, not informative: a prisoner's dilemma that enriches the platforms and bills everyone else.
  • Ads' one real virtue is universal access. Any alternative has to answer for losing it.
  • Each funder shapes the product. Ad money buys surveillance, VC buys growth that breaks federation, state money buys capture.
  • Sovereign infrastructure needs plural, staged, insulated funding: grants for the protocol, member fees and regional funds for the city layer, federation fees for the foundation.
  • Federation is not a feature, it is the capture defense. The funding model is the governance model.

Why ads are a net loss for almost everyone

The useful frame is the split between informative and combative advertising.

Informative adsCombative ads
What they doTell you a product exists, what it does, where to buyFight over existing demand: Coke vs Pepsi, every SaaS retargeting campaign
Share of spendA small minorityThe bulk
Welfare effectNet positive; reduces search costsPrisoner's dilemma; mostly cancels out

If both sides of a combative fight stopped, market shares would barely move and both would save billions. Neither can stop unilaterally, so the spend escalates. The cost is passed to consumers in prices, the rents are captured by the platforms (Google and Meta together take roughly half of global digital ad spend), and the winner's gain is temporary, lasting only until competitors match it.

The deeper critique, the one Scitovsky, Galbraith, and Tim Wu in The Attention Merchants all land on, is that advertising does not just allocate attention against existing preferences. It manufactures dissatisfaction to drive consumption. That makes the math worse: you are paying to be made unhappy with what you already have. Serious estimates put a majority of ad spend in the socially-wasteful column even when it is privately rational for each spender.

So the reflex to dislike ads is, in welfare terms, well-calibrated. But honesty requires naming the one thing ads get right.

Ads let poor and rich users reach the same product, with the money flowing in from advertisers. That universal access is the real virtue, and any alternative has to answer for losing it.

The funding question is the real question

Ads scaled because attention was monetizable at near-zero marginal cost. Reject them and you are not removing a funder, you are choosing a different one, and every funder shapes the product it pays for. This is the table I keep coming back to.

Funding modelHow it alignsWhere it breaks
AdvertisingUniversal free accessSurveillance, rent to the duopoly, manufactured demand
Subscription / SaaSIncentives align with the user who paysPaywalls fragment the commons; creates an access tier
PatronageWorks for individual creatorsDoes not scale to large shared infrastructure
Public / grant fundingSuited to commons-shaped goods; produces durable infrastructureSlow cycles, politically capturable
Foundation / endowmentDurable once capitalized (Mozilla, Wikimedia, Signal)Needs an initial capital injection
Open-coreFree core plus paid enterprise; fits sovereign infraTension over what stays free
Platform co-opMember-owned; no external rent extractionGovernance overhead; cold start is real

No single one of these replaces ads at the same scale and accessibility. The realistic stance is stacked: pay directly where you can, use commons infrastructure where it exists, fund the protocol layer with grants, and own the operating layer as a co-op. You do not fully exit the system. You starve the parts that extract the most rent and feed the parts that do not.

The thing being funded: local trades without the rent

I worked this out against a specific target because abstract ad critique goes in circles, and the failure modes of TaskRabbit, Thumbtack, and Angi are concrete. I covered the architecture in follow the rent; the short version is a platform that refuses every pathology of the incumbents.

Incumbent pathologyThe design that refuses it
20 to 30% commission rent per jobThe platform is not in the transaction; coordination and payment go direct
Relationship capture (TOS bans off-platform contact)Once two parties find each other, they are free; no lock-in
Reputation trapped in one siloPortable, signed reputation attached to a self-owned identity
Opaque ranking and paid placementFederated, geo-scoped discovery; friend-of-friend weighting, not ads

This is not theoretical. CoopCycle federates worker-owned delivery co-ops across roughly seventy European cities and replaces Deliveroo locally. Up & Go in New York returns 95% of fees to cleaning co-ops. Loconomics and Fairbnb solve versions of the same problem in their verticals. The model works. What none of them has done is unify it across trades with cross-border interop, which is where the protocol layer, and its funding, matters.

Funding it: layered, staged, insulated

The architecture is staged so no single funder can stall or capture the whole thing. Three layers, each with its own independent money.

three layers, three independent funding lines Foundation protocol stewardship, incorporated abroad federation fees + Horizon / Digital Europe City co-op the operating entity, one city at a time member fees + insurance + Interreg / ESF Protocol identity, federation, reference client NLnet + Sovereign Tech Fund a freeze in one line does not kill the others
The protocol is funded by grants, the city co-op by its own members plus regional setup funds, the foundation by federation fees and R&D grants. The point of three lines is that no single freeze, and no single captor, takes the whole system down.
PhaseWhat it fundsSource
0 · ProtocolSpec plus reference implementation. Sub-contractor stipends, no payroll.NLnet NGI Zero (Commons / PET), Sovereign Tech Fund. Apply per component.
1 · First cityStanding up one city co-op, organizer time, onboarding the first trades.Interreg (cross-border programmes suit a city near a border triangle), ESF+ via Local Action Groups, founding member fees, diaspora crowdfunding.
2 · OperationsOngoing running of a dense city instance.Provider membership fees (~60%), insurance partnerships (~15-20%), opt-in verification and escrow (flat fees, never commission).
3 · FederationProtocol R&D and stewardship across cities.Federation membership fees from city co-ops, Horizon Europe / Digital Europe.

What this funding model refuses

The refusals are as load-bearing as the sources. Each one is a door I am deliberately not opening, because each leads back to a pathology.

RefusedBecause
Venture capitalForces a growth model that breaks the federation thesis
Municipal capitalCompromises the capture defense; the city gets a seat at the table it should not have
Advertiser moneyRecreates the exact rent and surveillance the project exists to escape
Single-source dependenceOne frozen grant or one hostile funder should never be able to kill the system

Federation is the capture defense

This is the part that changes once you stop assuming good institutions. I take a high-corruption environment as the default, not because I am asserting it applies here, but because it might, and the asymmetry is stark: assuming capture when it does not materialize costs a little extra structure, while assuming it away when it does materialize costs the whole project. Design for the worse case and you are covered either way. Under that assumption federation stops being a nice-to-have and becomes the primary defense.

A single national board is one phone call from a connected person away from compromise. N independent city co-ops, federated by protocol, have no single point to capture.

If one city's instance is compromised, the others defederate and the protocol keeps running. That is the Mastodon and email pattern, not the platform pattern, and it argues for spending the first budget on the protocol rather than on any one instance. Whoever runs the first city matters less than making instance number five able to spin up independently and interoperate.

The state surface shrinks to transactional API calls and nothing more: the co-op legal form exists already, license verification is an API call to the professional registers, tax compliance is an integration, and last-resort disputes go to courts that already exist. None of that requires a governance role. Legitimacy and money come from outside the state instead: EU funding, trade associations, insurers who directly benefit from vetted providers, a civil-society anchor, and diaspora capital that sits structurally outside domestic patronage. Core IP lives in a foundation incorporated outside the operating jurisdiction (a Dutch stichting or Belgian AISBL), with the operating co-op as a licensee, and the protocol under a copyleft license so no one can fork it closed. It is the pattern that keeps Signal out of political reach.

The honest constraints

The thesis

The reason to care about all of this plumbing is that funding is not downstream of the mission. It is the mission's mechanism.

How you pay for a thing decides what it becomes. Ad money becomes surveillance and rent. VC money becomes growth that breaks federation. State money becomes capture. Sovereign infrastructure needs plural, staged, insulated funding, because the funding model is the governance model.

You cannot bolt sovereignty onto a captured funding structure later. A platform funded by a single national grant is centralized no matter how federated its code is, because the money is the single point. A platform funded by advertisers will surveil no matter what its privacy policy says, because the money demands it. Get the funding plural and insulated from the start, and the good governance properties follow almost for free. Get it wrong, and you rebuild the thing you set out to replace.

The first concrete move is the smallest one: the protocol grant. Identity and federation are a distinct, separately fundable component, and that is where this starts. One trade, one city, the protocol first, the federation only after the first city is dense.

Local trades is the clean version of the question, with a tidy member-fee line and a light regulatory load. For the same model under real load, applied to a patient-owned health record where the sickest can least afford to pay and the funder you accept could quietly decide who the app serves, see who pays for the patient's app.


I worked this out as a funding spine for a local-trades co-op, mostly as a way to pressure-test the idea, and it follows directly from follow the rent. The example is a vehicle; the model generalizes: any infrastructure that refuses rent has to fund itself plurally and outside single-point capture, or it backslides into the thing it set out to replace.