DePsy is a cooperative knowledge infrastructure for mental health, built on four interlocking layers. Each layer is real, named, and either operational or in active build. None of it is speculative.
DePsy's AI is not a proprietary model trained from scratch. It is a state-of-the-art open-source large language model (currently testing Mistral, Qwen, and DeepSeek model families under Apache 2.0 / MIT licensing) operated as a queryable surface over the DePsy knowledge commons. The model never owns the data. It reasons over data the contributors own — and cites every passage back to its source, with click-through to the original creator.
This is a deliberate choice. A model trained from scratch on survivor data would absorb that data into its weights, where it cannot be withdrawn, edited, or governed. By keeping the model and the data separate — using retrieval-augmented generation rather than wholesale training — DePsy preserves contributor sovereignty at the technical level, not just the rhetorical one.
DePsy's knowledge layer is being designed to deploy on Bonfires AI, the existing $KNOW Knowledge Network platform developed by DeSciWorld. Bonfires is currently operational across thirty-five active community deployments — including Hats Protocol, Regen Network, Secret Network, ETHBoulder, and MesoReef DAO — with eighty-eight thousand knowledge nodes and fifteen thousand indexed conversations.
DePsy's relationship with Bonfires AI is established and ongoing, with active collaboration toward a formal partnership for a DePsy-tuned deployment of the Knowledge Network architecture. DePsy is not building a Web3 system from scratch. It is the mental-health-reformation deployment of an architecture that is already operating, already federating across communities, and already returning $KNOW tokens to contributors based on retrieval activity. Each community node — a Soteria house, a peer respite center, a Hearing Voices circle, a Mad Pride collective — will operate its own sovereign Bonfire. DePsy federates across them.
DePsy participates in the Bonfires AI Knowledge Network as a cooperative. Each Bonfire that DePsy operates — the founding-cohort Bonfire, the federated community Bonfires anchored at Soteria houses and peer respite networks, and the meta-index that queries across them — earns $KNOW tokens based on retrieval activity from agents and applications querying its knowledge graph. The cooperative sets its own access pricing.
That earned $KNOW accrues to the DePsy cooperative treasury, not to individual contributors as automatic per-contribution payouts. How the treasury is distributed — to the contributors whose knowledge drove retrievals, to the physical respite network buildout, to platform operations, to the founding cohort, to new community Bonfires joining the federation — is decided by the DAO's governance. Contributors hold governance voice. Founding contributors hold founding voice.
This is deliberately not a "contribute and get paid" model. Per-contribution payout schemes are the failure mode of Web3 projects — they reward volume over value, drive token-extraction behavior, and collapse into speculation. DePsy's cooperative model rewards co-ownership over transactional contribution: contributors share in governance and in proportional treasury distribution as the cooperative determines, with the explicit expectation that a meaningful share of $KNOW value funds the physical infrastructure (Soteria houses, respite centers) the movement has needed for fifty years.
This is not a speculative new token. $KNOW is the existing unit of recognized contribution in the Bonfires Knowledge Network. DePsy participates in that network rather than creating a parallel one. Legal structure: cooperative, designed for compatibility with platform-cooperativism precedent (including U.S. cooperative statutes that recognize tokenized member ownership), and structured through governance design and contribution-weighted voting to prevent token concentration.
How $KNOW is actually earned, technically. Per the Bonfires AI documentation, each Bonfire's daily $KNOW allocation is determined by a Bonfire Score combining two factors: Gravitational Energy (the semantic alignment between the Bonfire's knowledge and what queriers across the network are actually looking for) and Retrieval Score (the frequency with which the Bonfire's knowledge is queried). In plain language: it is not "more contributions = more $KNOW." It is "more useful, semantically aligned knowledge that people actually need = more $KNOW." DePsy is designed to thrive in that economy because the survivor knowledge it holds is precisely the knowledge no centralized AI can offer and a growing population of researchers, clinicians, and communities urgently need.
Anti-gaming by design. The protocol applies degressive proportionality (mathematically related to Vitalik Buterin's √N framework) to weight diverse, uncorrelated contributions higher than coordinated or repetitive ones. This is a protective feature: the architecture actively rewards genuine variety of survivor voice over volume or coordination, which is exactly the property a movement-owned cooperative needs in order to remain a movement and not collapse into spam or capture. The full mechanic is documented at docs.bonfires.ai/bonfires/files/Knowledge+Economy/$KNOW and docs.bonfires.ai/bonfires/files/Knowledge+Economy/Knowledge+Network — readers who want to verify the math should go to the source.
The economics of Layers 1–3 fund the buildout of the physical layer: a global network of Soteria Houses, peer respite centers, Hearing Voices circles, and Open Dialogue hubs. The model — proven by Loren Mosher in the 1970s and replicated in research ever since — is not new. What is new is the cooperative funding mechanism that makes it possible at planetary scale, independent of government grants and philanthropic whims.
The single most important technical fact about DePsy is what it does not do. It does not scrape survivor content from the open web. It does not ingest YouTube channels without permission. It does not absorb published peer canon into proprietary weights. Every contribution to the DePsy corpus is explicitly granted by its creator, under the DePsy Founding Pledge — a one-page consent document that names what is being indexed, what use is permitted, what ownership stake the contributor receives, and the unconditional right to withdraw at any time.
The trust required to assemble survivor knowledge at scale does not survive centralization. It only exists inside cooperative governance. That is not an aesthetic preference. It is the architectural reason DePsy is possible and centralized AI alternatives are not.
Bonfires AI exposes four operational levels at which a Bonfire participates in the Knowledge Network: taxonomy-only (the Bonfire publishes only what topics it covers, not the substance), HyperBlog (curated long-form output drawn from the graph), conversational interface (queryable Q&A), and data rooms (full structured access for authorized parties). Each level affects discoverability, contributor exposure, and earning potential. Choosing the right exposure level for each DePsy Bonfire — and for each contributor's content within those Bonfires — is itself a governance decision for the cooperative. The default DePsy posture is contributor-protective: no contribution is exposed at a deeper level than the contributor has explicitly consented to in their Founding Pledge or subsequent governance vote.
DePsy is being designed to deploy on the Bonfires AI Knowledge Network architecture (developed by DeSciWorld), with whom DePsy holds an established working relationship and active collaboration toward formal partnership. The Bonfires platform provides the federated graph, $KNOW economics, and DAO governance infrastructure that DePsy will operate inside. Technical and governance leadership for DePsy specifically is being assembled across the alternative mental health movement and the decentralized-science community. The Bonfires AI Knowledge Network architecture is in active deployment, with the full $KNOW token economy scheduled for activation following the network's planned launch sequence. DePsy's deployment is calibrated to go live on the architecture as the network reaches full operation. For technical detail beyond this overview, see Bonfires AI's documentation at docs.bonfires.ai.