The problem with YouTube is not copyright law. It is that YouTube enforces far beyond it, because deleting your channel is cheaper than defending fair use in court. Hum a song on your podcast, strike, gone. Every centralized platform converges here: one company's risk aversion becomes everyone's speech policy.
Blossom solved half of this. Blobs are content-addressed, servers interchangeable, a takedown on one host doesn't touch the hash, clients already fail over. Three things remain unsolved: who pays for storage and egress, who keeps abuse off the shelves, and why a creator would publish here. Zaps have answered none of them. Zap volume is a rounding error, and people don't pay for what they can't first evaluate.
The solution
Invert the model: writing is free, reading is priced, and both run on one primitive, the sponsorship.
Tiers. A subscriber pays, say, ten dollars a month for 50 GB of reads, about thirty hours of HD, plus five sponsorship slots. A sponsored account reads a fifth of that for free, and gets slots of its own; accounts they sponsor read a tenth. One subscription carries thirty-one people. At wholesale egress under a cent per gigabyte this is profitable with room to spare. The product is not gigabytes, it is patronage.
Sponsorships are signed Nostr events: a grant from the sponsor, an acceptance from the sponsee. Because they are public and portable, any server can honor them, and a new server can join with the existing web of vouches instead of an empty user table.
Storage is a market. Each server keeps a blob only while its read revenue covers its storage and serving cost. Content people watch persists on many servers because it pays rent on each. Content nobody reads expires. Persistence stops being a promise an operator must defend and becomes a property of demand. Want something preserved? Be its paying reader.
Creators get a share. Each month a server pays part of what a blob earned, weighted by reader tier, to the pubkey that signed the video event, not to whoever uploaded the blob. Free discovery comes built in: video ships as HLS segments, each a hash-addressed blob, and manifests, thumbnails and low-res renditions stay free while full quality counts against quota.
Anyone can host. The whole thing is one binary next to a Blossom server: verify the auth event, compute the reader's tier from the sponsorship graph, meter the quota, sweep expiry and payouts monthly. It should run on every Umbrel and every spare NAS, turning idle bandwidth and storage into sats. With many small operators honoring one graph, serving fees race toward cost, and operators differentiate on policy and curation instead of margin.
The issues
Abuse. Free writing normally makes you the internet's dumpster. Here, write access is free of payment but never free of accountability: every writing key traces through sponsorships to a paying key. Slots are scarce and staked. If your sponsee uploads illegal content, your slots freeze; repeat failures cost your subscription. Not because you are guilty, because you are the underwriter. Invite trees have kept abuse out of gated communities for twenty years. No KYC anywhere: accountability is economic, not biographical. Public-facing operators can add hash-matching against known-abuse databases at upload; anonymous operators compensate with tighter trees. Policy lives at the edge.
Later
Eventually a server should be able to answer a request with a machine-readable 402, price, what it unlocks, payment pointers, so readers pay creators directly and clients comparison-shop servers. That is where serious creator revenue lives, and it deserves its own spec once the free-plus-subscription layer works.
What exists, what's missing
The stack is mostly there: khatru for relay hooks, blossom-server and route96, Flare and Plebs as video clients, NWC and Cashu for payments, BUD-01 authenticated GETs. Missing: the gate module (tier resolution, quota metering), the expiry-and-payout accountant, two one-page conventions (sponsorship kinds, HLS renditions as blobs), and client work: authenticated media requests, quota-aware playback, and a sponsorship UI normal people understand.
Numbers
- Bandwidth cost to serve 1GB: <$0.01
- Estimated monthly demand 90th percentile user: 50GB
- Sweet spot for what subscribers would easily pay per month if sponsoring many others with that: $10
- Monthly bandwidth paid at zero marginal profit: 1000GB
- Bandwidth needed for 3 tier, all 50GB fully sponsored, 5 invites each: (1 + 5 + 25) * 50GB= 1550GB!!
- Bandwidth needed for 3 tier, at 50GB/20GB/10GB, 5 invites each: (50 + 5 * 20 + 25 * 10) GB= 400GB
- Bandwidth needed for 3 tier, at 50GB/20GB/10GB, 10 invites each: (50 + 10 * 20 + 100 * 10) * GB= 1250GB!
Marketing
The invite system gets users into lower tiers. Power users get to experience the benefit and when they hit their quota, a strong incentive to upgrade.
By being invite-only, abuse is kept low but also a natural marketing chatter can evolve from people inviting their follows or others asking for invites.