#Identity and profilicity
Profilicity — A mode of identity in which a person forms and validates a sense of self by curating a profile and identifying with how it appears to a generalized observer. Not "fakeness"; it is how identity broadly works under broadcast media. The project does not try to escape it but to change its conditions. (§02, §02b)
Second-order observation — Observing oneself as one is observed by others — "being seen being seen." The experiential core of profilicity, and the thing a popularity metric hijacks. (§02b)
General peer — The abstract, generalized observer ("people") whose validation profilicity seeks. The project's aim is to shrink and warm this peer into known people and chosen communities rather than an anonymous crowd. (§02)
Sincerity / Authenticity — The two older identity technologies in Moeller and D'Ambrosio's account: identity through a committed social role (sincerity), and identity through expression of a true inner self (authenticity). The project explicitly does not market itself as a route back to "authenticity." (§02b)
Genuine pretending — A theory term for honestly acknowledging the performative nature of profile-identity, rather than pretending performance away. Useful in the profilicity account; not recommended as user-facing product language. (§02)
Contextual profile — A distinct, audience-appropriate face (name, photo, bio, visible channels) presented to a contact based on the context through which they reached you. Replaces the single flattened "main" account. (§03)
Profile sphere — The access-backed bundle behind a contextual profile: identifiers, keys, profile artifacts, visible channels, and default settings scoped to one audience context. (§03, §03)
Reach path — The route by which someone reaches a profile sphere: an invite link, QR code, exact username, or introduction. It is not a global people search. (§03)
Reveal grant — A later grant that adds profile material or channel access to a contact's existing card without announcing that a separate profile was revealed. (§03)
Link grant — A recipient-scoped assertion that two revealed spheres belong to the same owner for that recipient's local contact card. It is not a public equivalence edge. (§03)
Bounded anti-correlation — The limited privacy claim that unrevealed spheres should not be trivially linked through protocol-visible identifiers, keys, endpoints, artifacts, storage records, or recovery records. It is weaker than absolute unlinkability. (§03)
Recovery bundle — An encrypted, user-held package for restoring protected data and supporting an explicitly authorized device transition. It may include per-sphere material, manifests, local names and storage indexes. Restoration does not reactivate revoked devices, safely restore arbitrary old session state, or justify retaining every old content key. (§03)
Context (vs. circle) — An audience context that selects the profile and channels initially offered to a contact. Contexts are created through use rather than ranked relationship tiers; any suggested notification defaults remain subject to the recipient’s policy. (§03)
Profilicity ↔ authenticity tension — The standing evaluative lens: for any feature, ask which way it pushes on this axis and whether it re-introduces a visibility → popularity loop. Treated as permanent, not solved. (§02)
#Attention and network dynamics
Cumulative advantage / preferential attachment — The "rich-get-richer" process by which attention attracts attention, turning small early differences into steep, winner-take-all concentration. The hierarchy machine. (§02b)
Matthew effect — Merton's name for accumulated advantage ("to those who have, more will be given"); a synonym strand of cumulative advantage. (§02b)
Visibility → popularity feedback loop — The engine of the hierarchy machine: a visible popularity signal causes more attention, which increases the signal. The single highest-leverage thing the project removes. (§02b)
Herding (social influence bias) — Social influence on later judgments or behavior. In Muchnik et al., a manipulated visible early rating biased later ratings. Author-private feedback removes that tested public cue without claiming to eliminate all herding. (§02b)
Homophily — The tendency to connect with similar others; the structural seed of echo chambers. Flattening attention inequality does not, by itself, reduce it. (§02b)
Participation inequality (90–9–1) — The heavy-tailed pattern in which a small minority produces most contributions; activity-per-user traces a Zipf curve. Part of why some hierarchy is irreducible. (§02b)
Social-media prism — The distortion by which ranking-for-engagement amplifies the most extreme voices, making public life look angrier and more polarized than it is. (§02)
Generative social simulation — Embedding LLMs inside agent-based models to populate synthetic platforms whose rules can be varied — both prior evidence and a development tool the project intends to use. (§02b, §02b)
#Product mechanics
The two primitives — Conversations and channels. A conversation is something you say to someone; a channel is something you publish. They behave differently (reactions, receipts, notifications) because their audiences differ. (§04, §05)
Channel / sub-channel — The publishing unit; channels can nest (a DIY channel containing a sub-channel per project). Followed individually, at the level the follower chooses. (§04)
Ephemeral vs. permanent register — Ephemeral for "what I'm up to" (disappears, low-stakes); permanent for "what I made" (kept, findable). Ephemerality lowers the stakes of observation. (§04)
Excerpt-with-link — An attributed quote with commentary shared to the author’s chosen audience, subject to source disclosure and consent rules. It does not add a native repost mechanism or require a global discovery endpoint; authorized readers can still copy or aggregate material. (§06)
Author-private reactions — On a broadcast post, feedback is visible to its author rather than later viewers and does not notify other readers. In conversations, short-answer reactions may notify when they replace a reply, subject to recipient policy; ordinary emoji need not notify. (§06, §05)
Finite feed — A bounded, completing view ("you're all caught up") with no infinite scroll, pull-to-refresh-for-more, or algorithmic injection — the antidote to doomscrolling. (§06)
Declared (vs. inferred) ordering — Ordering set by the user's stated priority, never optimized against their behavior. The line between agency-preserving sorting and an engagement optimizer. (§06)
Discipline of omission — The principle that a feature is never added merely because it is possible; the prosocial path is made the path of least resistance. (§02)
Hostile-fork guard — Because the protocol has no concept of a global index, rebuilding a global ranked feed requires standing up separate indexing infrastructure — making the pathology expensive, not impossible. The main residual concern is modified clients and external aggregators, not a whole-protocol fork. (§06)
Crawler / aggregator resistance — The design constraint that TalkCo should not expose a convenient surface for harvesting public content, reconstructing social graphs, or inferring interests and political alignments at scale. (§01, §06)
#Safety and governance
Silent-by-construction blocking — Personal filtering and stopping future grants or delivery as applicable, without an explicit block notification. Shared-group changes, missing content or traffic may reveal exclusion; already delivered plaintext and keys remain. (§08)
Inherited blocking — Opt-in use of deliberately shared, depth-one block recommendations, without exporting private blocklists, automatic transitive propagation or person scores. Inspectable and overridable to avoid invisible exile. (§08)
Composable labelers — Opt-in, plural moderation services that label content/behavior in context (never a person-score); users subscribe to those they trust. Community safety without a reputation ladder. (§08)
Threshold-before-consequence — Corroboration or scoped facilitator review before ordinary contextual labeler consequences. Numerical counts do not establish independent reports; immediate protection need not wait for a quorum. (§08)
Restorative vs. protective tracks — Voluntary dialogue and repair for appropriate in-community friction, alongside prompt protective measures for credible threats. Protective action need not await restorative participation or a public verdict, and cannot promise unobservable exclusion. (§08)
Community account — An optional shared digital space for channels, records, resources and coordination. Sensitive access and custody powers are explicit; a full constitution, organizational chart or permission hierarchy is not required for ordinary work. (§09)
Technical steward — A person or group with expressly granted technical responsibility for devices, keys, backups, migration or continuity. Technical custody does not establish social leadership, content ownership or authority over physical resources. (§09)
Facilitator — A bounded support role for dialogue, repair, protective routing and scoped labeler review. Facilitation does not itself grant technical custody, private-content access or public rank. (§09)
Object-response gate — A legal/safety launch gate for bounded incident handling using the available object, manifest, locator, policy and controller evidence. A non-reading host may know only a locator and lease; intake must not demand unrelated private material or universal decryption. (§08, §12)
#Infrastructure and cryptography
Recipient descriptor — A protected record that binds an object’s exact encrypted representation and logical revision to its context, intended recipient key version, content-key information and accepted authentication profile. A storage locator alone supplies neither source authentication nor permission. (encryption research)
Authentication profile — The context’s accepted rules for authenticating content and control actions, including which evidence may be transferable. It comes from trusted context state; an incoming object cannot select a weaker verifier. (§10)
Cryptographic erasure — Loss of decryptability through destruction of the required keys, subject to every surviving key, descriptor, backup and recovery path. It differs from cooperating-host deletion and cannot make a recipient forget copied plaintext. (encryption research)
Threshold proxy re-encryption (TACo / Umbral) — The earlier collection’s umbrella label for related conditional-access candidates. Proxy re-encryption and threshold key release are distinct mechanisms; TACo and Umbral are not interchangeable implementations of every release policy. Operator, collusion, liveness and recovery assumptions require specific review. Refusing future release cannot revoke a key already delivered. (§10, encryption research)
Onion routing — Layered encrypted relaying that limits what individual relays learn about a path. Metadata guarantees depend on the observer model; timing and endpoint reuse can still correlate traffic. Session-style routes remain candidates within TalkCo’s transport comparison. (§10, SessionFAQ)
Session ID — A key-derived identifier used by Session. It is a carrier reference candidate, not TalkCo’s universal person identifier; TalkCo scopes identities and device records to contexts. (§03, SessionAccountIds)
Service Node / swarm — Session infrastructure terminology for service nodes and groups that buffer encrypted recipient traffic. It describes one carrier family, not TalkCo membership, content authority or long-term custody. Current network behavior remains an implementation-time check. (§10, SessionNodes, SessionSwarms)
Transport contract — TalkCo’s boundary above permitted delivery carriers. It defines traffic classes, scoped addressing, observer and metadata policy, unavailable-route behavior, expiry and storage boundaries. No carrier may silently downgrade a private context to an unapproved route. (§10)
Control plane — Coordination traffic such as invitations, reveal grants, configuration, acknowledgments and access requests, distinct from larger stored payloads. It still carries sensitive relationship and authority metadata and requires the context’s protection policy. (§10)
Forward secrecy — Protection of earlier encrypted messages after later key compromise, under the selected protocol and deletion assumptions. Retained plaintext, archive keys and backups remain separate exposure paths. (encryption research)
Erasure coding — Encoding data into redundant shares so a sufficient subset reconstructs it. Coding adds failure tolerance; it does not itself encrypt, repair, provide leases or erase content. (§11)
Hot tier — Better-connected storage or cache capacity for recent, frequently requested, safety-critical, or owner-declared important content. It is separate from cold mesh durability and must not become a hidden popularity system. (§11)
Seed/cache host — A reliable host class used for hot availability or redundancy support, distinct from intermittent ordinary friend devices. (§11)
Policy epoch — A version marker identifying an applicable access-policy state. It is distinct from a cryptographic group epoch and a custody-lease generation. A larger number alone does not authorize a transition; accepted causal and authority rules determine validity. (§10)
Sybil resistance — Mechanisms that make mass fake accounts costly without rebuilding the phone-number or global-identifier graph TalkCo avoids. Candidate approaches include invite scarcity, small payments, or accepting bounded abuse on a low-value free tier. (§11, §12)
Zooko's triangle — The observation that a naming system struggles to be simultaneously human-meaningful, decentralized, and secure; the constraint on human-readable usernames. (§03)
Signal-fork B→A — The earlier compatibility-bridge then independent-network working direction. It remains gated on service access, source/API reuse, privacy boundaries and migration evidence; it is not an adopted stack. (§10)