Each pattern below is stated with its intent, its mechanics, what it buys, and its limit. They compose, and they are a shared language rather than a mould: each member of the family combines them its own way and adds patterns of its own where its problem demands one. The full argument, with the economics and the standards each pattern rests on, is the pattern paper at https://1bsv.net/papers.html; bfinger is the first member to demonstrate many of them live.
The foundations are ratified open BRC standards throughout: self-proving objects (BRC-62/74/95/96), overlay state (BRC-22/24/88), key derivation and wallets (BRC-42/43/100), discovery (BRC-169/180), payments (BRC-29), keyed content (BRC-369), lookup markets (BRC-178), and the multicast object plane (BRC-148/149). Nothing here required a new consensus rule or a proprietary protocol; the design act is composition.
Intent. Use the chain for the one thing only a chain can do: put events in an order nobody can rewrite. Keep everything else off it, where storage is cheap, copies are free, and a retraction is real.
Mechanics. State lives in a chain of tiny mined outputs, the state tokens, each spending the previous one and each committing to the current record by a 32-byte hash. The record itself never appears in a block.
What it buys. Unforgeable ordering and history at a few hundred bytes per update, with double-spend protection doing the work that version vectors and consensus modules do elsewhere.
Limit. One chain per entity is a serialization point, by design. State that must change concurrently belongs to different entities with different chains.
Intent. Give data a name, a proof, and a revocation lever, without writing the data into a block.
Mechanics. The record rides an ordinary transaction whose fee is zero and whose lock time is a century away, so it can never confirm. Its transaction id is a 32-byte name for exactly those bytes. It spends an output of a parent transaction that is mined, so it is provable by ordinary SPV even though it will never be in a block.
What it buys. Content addressing, provability and retractability in one device, using nothing but transaction rules every verifier already implements.
Limit. A carrier is public to every host that carries its topic. Confidential content is encrypted before it enters a carrier, never after.
Intent. Make publication cheap for honest publishers and expensive for floods, with no rate-limiter service to run.
Mechanics. One mined transaction creates M small outputs; each published object spends exactly one. Every object is one proven parent plus one Merkle proof: constant size, independently verifiable.
What it buys. Structural rate limiting (no unspent output, no publication), constant verification cost per object, and a single point of retraction for everything a tree funded.
Limit. The tree's width is chosen ahead of need; a publisher that outruns it mines another tree.
Intent. Give high-rate streams chain time and completeness proofs without a miner fee per object.
Mechanics. Objects are numbered per producer and never mined. Once per interval, one anchor transaction commits a Merkle root over the interval's objects.
What it buys. An auditor proves the committed set for any interval against block headers alone, and a missing object is a numbered hole that any replica can fill. One fee per interval, regardless of rate.
Limit. Completeness is proven per interval, not instantly; the tail of the current interval is evidence in flight.
Intent. Let one small record vouch for a large body of content, revealed a member at a time or all at once.
Mechanics. A record commits to whole stores of further carriers by Merkle root. A linked store publishes its member list; readers rebuild the root and prove every member themselves.
What it buys. One mined output validates data amalgamated across many stores, and the record stays small however large the content grows.
Limit. A root discloses the store's size, and an unlinked store is unlisted rather than private; membership privacy needs an encoding that carries a secret.
Intent. Make advancing the state require more than holding a key.
Mechanics. Each record publishes the hash of a secret and withholds the secret. The next transition must reveal the preimage, which is checked against the previous commitment.
What it buys. A second factor on every transition: a published record proves its author held something it never showed anybody, and a stolen key alone cannot advance the chain.
Limit. The withheld secret exists in exactly one place, the publisher's own state, and is unrecoverable from the network by design.
Intent. Make retraction a mechanism instead of a request.
Mechanics. Spend the funding outputs the carriers depend on, and every carrier hanging off them becomes a double spend at once. Conforming hosts drop the records.
What it buys. One mined transaction retracts a record, or everything a tree ever funded, with no takedown process and nobody to petition.
Limit. Retraction removes future availability from honest hosts. It cannot un-copy what somebody already fetched, and nothing can.
Intent. Delete the distribution tier.
Mechanics. A publisher submits an object once, named by topic; the multicast object plane delivers it to every host subscribed to that topic, complete and verbatim, proofs intact. Loss is detected and repaired by the network itself.
What it buys. No fan-out servers, webhook farms, polling loops or reconciliation jobs; audience size stops being a cost the publisher engineers against. What the network moves is exactly what the reader verifies.
Limit. Delivery at the same moment is an architecture property, not physics; ordering across subscribers is never promised, and applications that need total order put it in the data, with a chain. And the language does not wait for the plane: every other pattern runs over plain unicast, publisher to host, with a good block-header source as the only hard dependency; the plane adds reliable distribution with global reach.
Intent. Make host choice an availability question, never a trust question.
Mechanics. Every subscribed host receives the same objects and admits them on their own validity. There is no protocol between hosts and nothing one host knows that another could not have received off the same plane. Readers who care ask several hosts and require agreement.
What it buys. Interchangeable hosts, load balancing that is safe to do simply, and a market where hosts compete on availability, latency, retention and richer questions, never on custody of the data.
Limit. The cost of a bad host is a refusal and a retry; disagreement between replicas is surfaced to the reader, who decides.
Intent. Keep participation free, and keep it anonymous.
Mechanics. A price attaches to a question class, never to part of an answer. The base classes are free on every conforming host; a priced class may not be a superset of a free one; a question carrying an unknown member is refused, so no spelling manufactures a priceable alias of a free question.
What it buys. A directory that can never become unreadable, and anonymous reads at the floor: a host that charged for the base answer would have to identify the payer. The floor is a floor, not a ceiling: richer questions, history, exports, availability proofs and keyed content are all open to pricing above it, and a paid read is attributable by nature, which is exactly why the anonymous base answer is pinned beneath it.
Limit. A restriction is a property of bytes, and only encryption produces one; a service fee is a property of a socket. A priced lookup sells availability and convenience, never exclusivity, and must never be presented as access control. Paid content that must stay paid is keyed content: the payment releases the key.
Intent. Keep the delivery meter and the money separate, and leave every payment topology open.
Mechanics. Any verified identity key is already a payable destination. The simplest leg derives a fresh output from it, pays point to point, and tells the recipient how to claim it. Richer shapes build on the same substrate: a payment can travel inside a record and settle when the recipient claims it; a bounty can be published as a record that competing providers race to claim, singly or as a series funding ongoing service; a payment can split across many receivers; and an interactive payment flow can be settled by an overlay itself, its topic rules judging the outcome at every host identically.
What it buys. No address reuse, a payment graph that is not a broadcast artifact unless the design wants it to be, and revenue seams at every edge that faces an audience: priced question classes, subscriptions, bounties, markets across competing hosts. The delivery meter never bills a payment as a payment.
Limit. A broadcast shows everyone the same bytes, so a payment that must reach one party is derived per recipient, and a multi-party settlement puts its judging rules in the topic, where every host applies them alike. Notice delivery is a channel of its own; automating it is the message-box member's job.
Intent. Never let the party who served the answer also supply the yardstick.
Mechanics. Every object carries its Merkle proofs; every reader and every host checks them against block headers it received itself, from a header lane, never from whoever answered the lookup.
What it buys. The chain of "our API says so" becomes a chain of arithmetic anyone can rerun, and a host that edits a record produces something that fails the check rather than something believed.
Limit. SPV proves inclusion and ancestry; a reader with no host it trusts still needs an index to prove a spend, and designs say so rather than implying otherwise.