DeSci Labs

Overview

DeSci Labs is a Swiss software company building the base infrastructure for decentralized science – the open, content-addressed layer beneath the field's publishing and funding platforms. Where BioDAOs such as VitaDAO or ResearchHub coordinate who funds and reviews research, DeSci Labs builds the plumbing that makes a research output durable, versioned, and machine-readable in the first place.

It is the first infrastructure and standards organization in this directory rather than a funder or a patient community: its products – DeSci Nodes, the dPID identifier protocol, and the DeSci Codex network – are the components the decentralized-publishing pillar cites as the tooling under everything else.

The problem it targets

DeSci Labs was founded on a diagnosis of the traditional scientific record: results are frozen into flat PDFs that strip out the data, code, and provenance, then locked behind a publishing cartel at margins that dwarf most of tech. CEO Philipp Koellinger frames the deeper cost as a “replication crisis that is derailing scientific progress” – a problem the wiki treats at length in the reproducibility-crisis page.

Two failures follow from the PDF-as-unit-of-record model. Citations point at link-rotting URLs rather than the exact bits that were reviewed, so the literature quietly drifts out from under its own references; and the underlying data and code, severed from the paper, cannot be independently re-run. DeSci Labs' answer is to change the unit of publication itself.

Research objects: DeSci Nodes & DeSci Publish

The core move is to publish a research object instead of a document: manuscript, datasets, code, figures, and metadata bundled into one versioned, independently-addressable unit. DeSci Nodes (also shipped as DeSci Publish, billed as “the first open-source, peer-to-peer platform for scientific publishing”) is the application researchers use to assemble and release these objects, with private drafts, licensing, and transparent version history built in.

The engine is open source: the desci-labs/nodes repository describes Nodes as “secure PID containers for interoperable research objects.” Around it sits a developer surface – Codex-Lib for low-level protocol access, Nodes-Lib and a Nodes CLI for automated publishing – so other DeSci applications can publish to the same network without rebuilding the storage layer.

dPID: decentralized persistent identifiers

When a Node is published it automatically receives a dPID (decentralized persistent identifier). Unlike a DOI – a mutable pointer administered by a central registrar – a dPID is built on content identifiers (CIDs), the cryptographic fingerprints used by IPFS, so “the same logical object always maps to the exact same sequence of bits.” That guards against content drift and lets a reader cite an individual component of a paper, not just the whole.

dPIDs read like short, human-friendly filepaths with version and history control, resolvable through the dPID Browser or an open-source HTTP dPID Resolver that bridges the conventional web to decentralized storage. Notably the protocol is a split responsibility: per dpid.org it is “Engineering by DeSci Labs & Governance by the DeSci Foundation” – the company writes the code, a separate non-profit stewards the standard.

DeSci Codex & the Open State repository

DeSci Codex is the open protocol beneath the app – “a decentralized network for publishing and preserving scientific research objects.” Content is content-addressed and securely replicated across multiple nodes via peer-to-peer resolution, so a published object survives independent of any single service provider. Institutions can run Codex-Node backend services to participate in that replication and add durability to the network.

The stated payoff is research that is durable, reusable, and AI-ready: because every object carries its data, code, and an auditable provenance chain – who published what, when, and what changed – results become both independently reproducible and directly machine-readable, the container the flat PDF never was. It is the same open-record shift the rest of DeSci makes with IP-NFTs, applied to the publication itself.

SciWeave: an AI assistant for the literature

DeSci Labs' newest product reads the scientific record rather than publishing to it. The company describes SciWeave as “an AI research assistant that helps users find, analyze, and summarize academic studies with accurate, citation-based answers”. It ships as a mobile app on iOS and Android (Google Play listing) and as an MCP server, which its repository describes as a “Claude connector for research paper search and AI-powered answers”; that repository was created on 30 March 2026 and last pushed to on 2 August 2026.

The company's main claim for SciWeave concerns citations. In a post published on 2 August 2026, Koellinger and Ogban Ugot report a test on 100 randomly drawn biomedical and neuroscience questions from ScholarQABench, the citation-grounded question-answering benchmark introduced with Asai et al.'s OpenScholar work in Nature. SciWeave cited no paper that does not exist; GPT-5.2 invented 59% of the papers it cited, and Claude Opus 4.6 invented 62%. The post states two conditions on that result: DeSci Labs tested its own product, and published replication scripts; and it called both comparison models directly through their APIs with no retrieval tools, so the test compares a system that searches a paper corpus with models answering from memory. It measures what retrieval does to invented citations, and it does not compare SciWeave with another retrieval-based assistant.

Company and Foundation: engineering vs. governance

DeSci Labs deliberately separates the venture-funded engineering company from the neutral steward of the standards it depends on. The DeSci Foundation is a non-profit registered in Canton Zug, Switzerland (CHE-389.450.872, subject to annual Swiss audit) whose mission is to “explore and support improvements in the scientific ecosystem” and to “enfranchise stakeholders” in the emerging field.

The two are organizationally distinct but aligned: Koellinger is the Foundation's president as well as DeSci Labs' CEO, and co-founder Christopher Hill sits on its five-member Executive Board, which works alongside an advisory board of 15+ from academia, blockchain, and open science. Keeping the identifier standard's governance in a separate non-profit – rather than inside the company that could later be acquired or wound down – is a bet on credible neutrality for infrastructure the whole field is meant to build on.

Funding, team & place in the ecosystem

DeSci Labs closed a $6.5M seed round on 29 August 2024, backed by Stride VC, OSS Capital, Protocol Labs, Accomplice, HV Capital, and Holtzbrinck Digital, with the capital earmarked for a “traceable digital scientific record” and the DeSci Publish launch. The team of roughly a dozen engineers and scientists is led by Koellinger (a behavioral-genetics researcher with 17,000+ citations, formerly of the University of Wisconsin–Madison and Vrije Universiteit Amsterdam), Hill (a neuro-economist), and Iman (a software engineer previously at Palantir).

Within the wider field DeSci Labs is the layer under the platforms: ResearchHub pays for peer review, Molecule and Bio Protocol route funding into BioDAOs, and DeSci Nodes plus dPID give all of them a durable, content-addressed way to name and preserve what the science actually produces.

How Caper approaches this

DeSci Labs is not a governance protocol – it publishes research, not treasuries – but the two share a load-bearing idea: a decision, or a result, should be a durable, content-addressed record that anyone can audit, not a claim you take on trust. A dPID pins the exact bits that were reviewed so a citation can't rot out from under itself.

Caper makes the equivalent move for what a DAO does with money: every treasury action is a typed proposal – including a PAYOUT that transfers from the treasury, an INVEST that buys another caper's token and a DIVEST that sells one back – recorded and executed on-chain rather than settled in a back channel. A proposal that moves no money takes the same typed path, such as a METADATA proposal that rewrites the mutable metadata of the caper's own token. And it echoes DeSci Labs’ second bet, splitting the neutral standard from the company that builds it – though in Caper the split lands a layer lower down.

The bonding curve’s cap is pinned by the state package, which is published immutable and which the contract source says no logic version can violate; the curve’s slope constant and the whole fee schedule sit in the swappable logic package, and moving either takes an UPGRADE proposal in the $CAPER caper rather than a vote in the caper being repriced. So a caper’s own ballot reaches its treasury and nothing else – not because the market is beyond governance, but because it is beyond that caper’s governance. A research collective that publishes its outputs through DeSci Nodes could fund that work through exactly these typed, auditable proposals. (Caper mechanics verified against contracts/core/src/caper_dao.rs and contracts/logic/src/lib.rs.)