01 · Project Overview

#Aims and reasoning

#How a "social network" became "social media"

The thing people mean when they say they hate social media is rarely a particular feature. It is a feeling: that the medium leaves them more performed, more compared, and more drained than connected. That feeling has a structural cause, and it has two parts worth naming before anything else.

First, the platforms quietly changed what they were. A social network once meant the people you know and the ties between them. Social media is a different thing — a publishing and broadcast medium, ranked by an algorithm, in which you address a vast anonymous audience and are implicitly ranked against everyone else addressing theirs. The social network was the bait; the media machine became the product. (The full account is in §02.)

Second — and this is the part most often missed — the features that produce these dynamics were mostly added because they could be, not because anyone reckoned with what they do to how humans relate. A public like count, an infinite feed, a one-tap reshare, a visible follower number: each is an obvious thing to build, each ships a measurable engagement bump, and each quietly bends behavior toward comparison and status-seeking. The harm is rarely intended. It accumulates from a thousand locally reasonable additions.

TalkCo is built the other way around. Every feature is judged first by what it does to the way people relate to one another, and the discipline of deciding not to build something is treated as a primary design tool rather than a failure to ship.

#Warm messaging, and a place to publish — for the people you actually know

TalkCo is two things at once: a prosocial, private messenger, and a place to publish your own content — deliberately scoped to your social network (the people you know, the people they know, your mutuals) rather than the global public. It is, in short, a warm alternative to social media that is specific to your actual social network instead of the world's.

A place for developing human connection, not human capital. That line names the design orientation: TalkCo should make relationships easier to tend without turning those relationships into audience, reach, leverage, or quantified standing.

The capability that makes this coherent rather than just another inbox is graceful contextual identity: the ability to present different, appropriate faces to different parts of your life without collapsing them together or resorting to a separate app for each. People today keep close friends on one messenger and loose or professional ties on another because each app is, in effect, a different audience. Hold context gracefully and a single warm space can hold far more of a person's relationships without the dread of the wrong audience seeing the wrong thing. The detailed treatment is in §03.

#What you actually do here

There are two primitives, deliberately kept distinct:

• Conversations — private one-to-one and group messaging. This is the messenger, and it is first-class, not a bolt-on.
• Channels — the unit of publishing. A person keeps channels for different parts of their life (friends, a hobby, the projects they are working on, professional updates), and shares text, images, video, and files into them. Channels can nest (an overarching DIY channel with a sub-channel per project), can be long-lived or ephemeral, and are followed individually rather than as one undifferentiated stream. (Detail: §04.)

Around those two primitives:

• Publishing without a website. The thing the open web never made easy for non-technical people — durable, structured content under your own identity — without running any infrastructure.
• Contextual profiles. What someone sees of you depends on how they reached you (a QR code, a link, a username). (Detail: §03.)
• Decentralized and user-controlled. Identity, encryption, hosting, and access control are designed to avoid a central custodian. (High-level: §10, §11.)

#Designed with the consequences in mind — what TalkCo refuses to be

These are constraints held on purpose, not gaps waiting to be filled:

• Not a global feed or engagement platform. No planetary, algorithmically ranked stream optimized for time-on-app. (§06.)
• Not an exploration or aggregation substrate. TalkCo is not built for crawling, harvesting, public discovery, or cross-network social-graph analysis. The core threat model includes corporate and state actors trying to infer relationships, interests, and political alignments from graph structure.
• Not status-driven. No public follower counts, like counts, vote scores, karma, or virality metrics functioning as competitive signals. (§02, §02b.)
• Not growth-optimized. As a non-profit effort, the platform does not instrument or incentivize maximal engagement; where the project needs health signals for its own learning, they measure connection quality, not attention captured. (§12.)
• Not a surveillance or data-broker substrate. Context boundaries, metadata minimization and non-reading custody constrain every candidate architecture.

Underlying these constraints is the meta-discipline already stated: a feature is never added merely because it is possible.

#Protocol and architecture

#What sits underneath

Sharing an implementation does not merge audiences or histories. A host can provide capacity without joining the community or receiving its content keys. The product remains two primary surfaces; optional shared records reuse the same boundaries (shared-core study).

These are implementation hypotheses, not a dependency lockfile. The older Session/TACo/Signal-lineage direction remains visible in §10, under explicit comparison with the newer research. A low-complexity prototype does not automatically meet the privacy profile required by a vulnerable community.

#Open Threads

The overview introduces no separate decision queue. Product scope and release sequencing live in §12; exact stack adoption remains in architecture lane #1.