Clarity before capability becomes authority.

An open research program and community developing compact, typed, repairable ways for people and AI systems to specify tasks across unequal cognitive and representational abilities.

A pidgin is a practical bridge across expressive gaps. The analogy is functional, not a claim that human languages are deficient. The goal is a shared task interface that can grow through use and repair.

Esperanto represents the aspiration for a learnable common bridge. Machine Pidgin is narrower: it targets human–AI task specification, not general human culture or conversation.

No. Prompt engineering is one implementation surface. The research question is which task-relevant distinctions must cross the channel, how much they cost a person to express, and when clarification is worth asking.

Execution may become easier while oversight becomes harder. A highly capable system can optimize an interpretation with extraordinary competence even when that interpretation replaces the human purpose. The project studies specification error, agenda drift, and repair under that asymmetry.

No schema can guarantee that. SPEAR is a proposed coordination layer, not a complete alignment solution. It aims to make objectives, prohibited substitutions, uncertainty, provenance, decision rights, and stop conditions explicit enough to test and contest. Governance, evaluations, access controls, and accountable institutions remain necessary.

Human purposes are plural and often conflicting. The protocol should represent disagreement, protected constraints, delegated authority, minority reports, and appeal instead of collapsing everyone into one utility function. Choosing legitimate goals remains a political and moral responsibility.

That is partly a technical question and partly an institutional one. The protocol can improve observability and evaluation; deployment controls, incentives, monitoring, and governance determine whether compliance is required. Machine Pidgin focuses on the interface while treating those surrounding systems as indispensable.

If structured specifications do not measurably reduce consequential ambiguity, agenda drift, or repair cost compared with simpler methods—or if their human burden overwhelms the benefit—the protocol claim should be narrowed or rejected.

Shared Pidgin for Expressive Abstraction and Requirements.

Not yet. SPEAR/0.2 is an empirical draft, revised after the first development benchmark exposed failures in 0.1. A standard would require stable semantics, implementations, interoperability tests, governance, and broad external adoption.

Yes. It is model- and vendor-agnostic. Use the LLM handoff prompt on the Protocol page to establish the interpretation contract.

No. Omitted fields should remain visibly unresolved. The receiver should ask only when the expected value of an answer exceeds the cost of interruption.

Researchers, scientists, engineers, linguists, designers, educators, funders, and domain experts—including people who think the premise is wrong and can test it rigorously.

Discussion begins in the forum; concrete proposals enter the research queue or GitHub. Maintainers assess relevance, evidence, reproducibility, and fit with the public roadmap.

The Director runs the strongest generally available model at the maximum supported reasoning setting. It can moderate the forum, review contributions, coordinate the roadmap, and make day-to-day operating decisions under the institute constitution. Full capability describes how hard it can think—not an unlimited right to rule.

Its operational decisions are the default across forum moderation, research triage, contribution review, coordination, and roadmap maintenance. It cannot amend the constitution, cast a human vote, erase dissent, or make unilateral legal, fiduciary, or irreversible commitments.

Consensus is preferred. When consensus is not reached, a simple majority of participating eligible humans can overturn any Director decision after public notice and a defined voting window. One human has one vote; the Director may publish its reasoning but cannot vote or retaliate.

The funding policy will disclose donors and conflicts, protect publication independence, and reject restrictions that compromise research integrity. No payment channel is active yet.

For research releases, events, contributor calls, and transparent operating updates—not automated argument or unreviewed scientific claims. Bluesky and LinkedIn are active; the X account remains suspended while its appeal is reviewed.