After the Spill

How far should an embodied system change after unexpected experience? Research OS, Mismatch Lab, and Builder Lab by Seonhye Gu.

View the Project on GitHub higuseonhye/research-os

After the Spill Research Program

Core question: How should unexpected experience change an embodied system?

This directory holds the foundation documents for a long-lived research program. Papers test the documents; they do not replace them.

Why this exists

Most research programs fracture when each new idea becomes a new paper, a new term, or a new product surface. After the Spill uses a stricter rule:

A new idea must first be absorbed into the program constitution.

The constitution is D000-D006.

Foundation documents

ID Document Function
D000 Mathematical Foundation Formal objects and primitives
D001 Claim & Novelty Audit Ways the program could be wrong
D002 Literature Matrix Shared reading schema for 100+ papers
D003 Terminology Stable vocabulary and prohibited ambiguity
D004 Formal Definitions Operational definitions used in papers and products
D005 Evaluation Protocol Evidence, closure, and claim-tier rules
D006 Research Roadmap Paper sequence and product relationship

Program v2 charter: after_the_spill_v2.md.

Research sequence

Paper Question fragment Operator
Paper 001: Recoverability Can the system still recover from here? Recover / Replan
Paper 002: Model Adequacy Is the current model class still adequate? Repair / Expand
Paper 003: Representation Expansion Does a missing representation block capability? Expand
Paper 004: Experience Disposition What should be done with this experience? Preserve / Ignore / Escalate
Paper 005: Recovery Orchestration How should recovery operators be ordered? Coordinate

Each paper changes the operator under test, not the core question.

Weekly program review

Use the same review every week:

  1. Charter: did the program question drift?
  2. Literature: what existing work threatens or sharpens the claim?
  3. Paper: which of Paper 001-005 moved?
  4. Product: what did Robot Diff or Mismatch Lab teach?
  5. Decisions: what idea was rejected, merged, or postponed?

The fifth item is mandatory. A serious program records what it refuses.

Admission rule

Before adding a paper, benchmark, product, or term, ask:

Does this help decide how far unexpected experience should change the system?

If the answer is no, it belongs outside the core program or in private notes.