Skip to content

How Humanities Superpowers differs from Superpowers

Humanities Superpowers is inspired by the idea of composable agent skills, but it is not a direct translation of software-development practices into academic prose.

obra/superpowers orientation Humanities Superpowers orientation
software development methodology humanities research methodology scaffold
test-driven development claim–evidence–counterargument verification
debugging program behavior diagnosing conceptual slippage, interpretive overreach, and category errors
implementation plans argument maps and research boundaries
code review manuscript review and reviewer simulation
tests before completion source, terminology, argument, privacy, and submission gates
executable correctness accountable but contestable interpretation

What is retained at the level of principle

  • systematic rather than ad hoc process;
  • evidence before completion claims;
  • reusable, composable skills;
  • explicit conditions for invoking a skill;
  • stop conditions when required evidence is missing;
  • verification as a distinct final act.

What is deliberately changed

1. No humanities equivalent of a passing unit test

Interpretations cannot usually be reduced to binary tests. The framework therefore checks traceability, proportionality, counterevidence, conceptual consistency, and disclosed uncertainty rather than declaring an interpretation objectively proven.

2. Close reading is not data extraction

The agent must connect an interpretation to specific formal, rhetorical, textual, visual, or material features. It must also record plausible rival readings.

3. Citations are not decorations

A citation must exist, be correctly identified, and actually support the claim attached to it. Bibliographic completion is not allowed through guessing.

4. The framework resists ghostwriting

structuring-humanities-argument organizes researcher-provided claims, notes, and evidence. It marks missing warrants and unsupported transitions. It does not authorize the agent to impersonate the researcher's original scholarly judgment.

5. Uncertainty is an output

When verification fails, the correct result is an unresolved-item list—not fluent filler.