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

Mismatch Lab

Public Lab · external entry point for Research OS
Status: v0.1 spec · pilot design · not yet deployed


What is Mismatch Lab?

Mismatch Lab is the public-facing spec for understanding robot behavior — starting with Robot Diff, not failure analysis.

Surface Role
Research OS Evidence engine — prereg, protocols, Paper 001/002, negative controls
Mismatch Lab Open spec — demos, benchmark, explorer design, SDK sketch, pilot invitation

Commercial operations (hosting, contracts, SLAs) are not in research-os. See vision narrative for long-term research framing only.


One-line thesis

What changed between these two robot runs — and when does that change mean the model needs more than retuning?


Documents

Doc Purpose
v0.1 spec Product spec · 4 cases · Diff UI · 4–6 week build plan
Diff Explorer (live demo) Static interactive wireframe · GitHub Pages
API schema mismatch.diff · mismatch.analyze · report JSON
Homepage v0.1 Hero · CTAs · demo copy · pilot waitlist
Benchmark spec Adequacy benchmark · negative controls · scoring
Product overview v0.1 Public-safe problem · product · evidence summary
Public scope What belongs / excluded from this folder
Vision narrative v0.2 Long-term research framing · not a company roadmap

User-facing verbs

Simple public entry · adequacy as research-backed insight (not the hero).

User sees Internal engine
Replay Trajectory visualization
Diff Aligned difference map
Explain Timeline + salient moments
Discover Interesting episodes in corpus

Premium insight (not front-page): Adequacy — persistent mismatch that repair cannot absorb.


Paper 002 (EXP-SURG-003) provides scientific evidence for the adequacy decision rule.
See Paper 002 hub · confirmatory spec.


Updated 2026-07-29