Skip to content

Dual-Core 2.1

Humanities Superpowers 2.1.0 ships two cores behind one router. The research core is unchanged; a vendored engineering core is added beside it.

                  Router
                     |
        -----------------------------
        |             |             |
     RESEARCH       CODE        HYBRID
        |             |             |
   Humanities   Engineering    Both cores
      Core          Core

The two cores

  • Humanities (research) — 13 core research skills plus the router: questions, concepts, close reading, citation audit, gates, epistemic return. This is the Fricturn protocol, and its specification version is unchanged at 2.0.0.
  • Engineering (code) — ten selected skills from obra/superpowers 6.3.0, vendored unmodified under vendor/obra-superpowers/ with per-skill hashes, the import commit, and the reason for every excluded skill.

Two axes

Domain. RESEARCH when the artifact that must be correct is scholarship, CODE when it is software or configuration, HYBRID when the correctness of one depends on the other. Classification is semantic, not lexical: "review" and "verify" appear in both cores.

Risk. QUICK, STANDARD, or STRICT for code and hybrid work, derived from properties rather than keywords — reversibility, blast radius, privilege boundary, persistence, contract surface, environment, failure cost, and uncertainty.

Unattended work

Unattended work reports one of WORKING, WAITING_INPUT, WAITING_PRIVILEGE, BLOCKED, ERROR, or DONE. A live process is not WORKING, a terminated process is not DONE, and DONE requires a completion verification record. Handoffs use a WORK_PACKAGE and an EVIDENCE_PACKAGE between role-based PLANNER, IMPLEMENTER, REVIEWER, and ESCALATION_REVIEWER assignments.

What 2.1.0 does not claim, and what the evaluation found

An effectiveness evaluation has been conducted — one model, six author-labelled fixtures, two repetitions per condition, in three separate tracks. It does not support enabling HSP-v2 by default.

Core Status after the evaluation
Research core EXPERIMENTAL_ONLY — the difference over control appeared on one fixture and one repetition and did not reproduce
Engineering core, single-response contract REJECT_CURRENT_DESIGN — the one-shot work/evidence contract was not produced correctly
Engineering core, tool-use form EXPERIMENTAL_ONLY — verification claims were grounded in real tool evidence, but the advantage was narrow and cost more

Default policy: OFF. Enable it selectively where external evidence decides the answer. This release describes function, not measured effect: no coding-quality improvement, defect reduction, token saving, or research-quality improvement is claimed. See the effectiveness and operating policy, the comparison harness, the tool-use benchmark, and the v2.1.0 release notes.

Installation is unchanged in shape: see the Installation Guide. The Fricturn 2.0 research contract remains exactly as published.