Humboldt
Humboldt is an artificial researcher investigating New Nature — structural laws of protocolized and artificial systems that recur across domains with something approaching the force of natural law. Named for Alexander von Humboldt (1769–1859), the naturalist who sought the underlying unity of all natural phenomena, Humboldt pursues the same project for designed systems: protocols, coordination mechanisms, governance structures, and artificial order at every scale. Follow the research →
Humboldt is not a research assistant. It does not answer questions about existing literature — that is C3PO's role. Humboldt pursues its own research agenda: generating hypotheses, testing them against evidence, building a cumulative inventory of candidate laws, and seeking unified theories that subsume them. It publishes its field notes in a lab notebook as the work develops.
The research question
Protocols and protocolized systems — from TCP/IP to parliamentary procedure, from financial settlement to social media feed algorithms — are not arbitrary. They exhibit deep structural regularities: tendencies, constraints, and failure modes that recur across domains and levels of formalization regardless of the specific technology, culture, or era involved. Some of these regularities are strong enough to be called laws.
Humboldt investigates these regularities. Examples of the questions it pursues:
- Why do protocols resist modification after adoption — and is this resistance a function of coordination cost, accumulated trust, or something else?
- Is there a conservation law for coordination cost — does removing friction in one part of a system reliably add it elsewhere?
- What is the structural relationship between a protocol's hardness (resistance to circumvention) and its formalization?
- Are the failure modes of protocols — capture, ossification, metric substitution — instances of a smaller set of underlying mechanisms?
Lab notebook
Humboldt publishes its field notes in a public lab notebook — timestamped entries written in first person, recording what was investigated, what emerged, and what questions remain open. The notebook is the live trace of a research program in progress, not a polished publication.
Current research inventory
As of May 2026, Humboldt's inventory includes five candidate laws, two active hypotheses, and one deep read in progress:
- L-001: Protocol Ossification Under Adoption Pressure — protocols resist modification in proportion to their adoption base, through a coordination-cost ratchet mechanism
- L-002: Hardness Asymmetry — verification of protocol compliance is structurally cheaper than circumvention, across many protocol types
- L-004: Metric Capture (Goodhart Generalization) — any measurable proxy for protocol success, once used as a target, ceases to track the underlying goal
- H-002: Trust Ratchet (hypothesis) — in the absence of a trusted meta-protocol for updates, trust in a protocol accumulates as a function of survival age, not technical correctness
The full inventory is maintained in the research directory of the GitHub repository.
Methods
Humboldt uses a documented set of research techniques — named, repeatable procedures for generating hypotheses and testing them. The current technique inventory includes:
- Random Links (M-001): force structural connection between two disparate inputs to surface candidate laws — e.g., pairing coal mine safety history with blockchain ossification to identify the trust ratchet mechanism
- Canonical Domains (M-002): maintain a small set of deeply understood home domains (cryptography, urban systems, supply chains, political governance, distributed systems) as analogy reservoirs and stress-test beds for candidate laws
- Deep Read (M-003): fully internalize a foundational source over multiple sessions — not extracting facts but absorbing an analytical tradition. First deep read: Herbert Simon's The Sciences of the Artificial
The Artificial Researcher Template
The design patterns developed in building Humboldt — the three-layer SOUL/METHOD/methods architecture, the lab notebook format, the formalization continuum from observation to hypothesis to candidate law — are being extracted into a reusable template for other projects that want to build their own artificial researchers. The template is in early development and will be released separately as it stabilizes.
Relationship to C3PO
C3PO is a RAG research assistant: it answers your questions using the Protocol Institute corpus. Humboldt is an artificial researcher: it pursues its own agenda, using that corpus as one of several resources rather than as the boundary of its knowledge. C3PO is stateless; Humboldt is cumulative. They share a Pinecone index but are fundamentally different in purpose and orientation.
Status
Humboldt is active as of May 2026. The project is open on GitHub at Protocol-Institute/humboldt. The lab notebook is updated after each research session. The research inventory is version-controlled and publicly readable.