What is Vibe Science?

Vibe Science is the applied study of how weak human signals and machine inference become validated decisions under uncertainty. It formalizes the discipline between intuition and metrics using the VSV (Vibe → Signal → Validation) loop.

What is the VSV Loop?

The VSV Loop is a three-phase methodology: Vibe (human perception, intuition, discomfort, excitement), Signal (weak data amplified by machines), and Validation (structured experiments, often co-designed with AI). The novelty is that machines enter before certainty, not after.

How is a Vibe Scientist different from a Vibe Coder?

Vibe Coding is tactical prototyping. Vibe Science is systems-level inquiry. A Vibe Scientist designs and validates hypotheses at the boundary between human intuition and machine inference — creating probes, instruments, simulations, and falsifiable narratives.

Who is already doing vibe science work?

Many existing roles already practice aspects of vibe science: Chief Product Officers synthesizing weak signals, CTOs making architectural bets under uncertainty, Heads of Design holding ambiguity, Chiefs of Staff detecting organizational drift, UX Researchers clustering qualitative data, and early-stage founders operating on pattern recognition daily.

Experiments

Generated video cannot hold still, and the numbers we quote do not travel

Running

Context

August 2026 measurements on a single Jetson AGX Orin. A real locked camera holds still for 28.8 minutes. Generated footage we measured does not stay still longer than 4 seconds on the first architecture (86x gap). Feeding a video model its own output collapses it, and that law reproduced on a second architecture, but the collapse point moved by a factor of five. SSIM reports 0.9998 after destroying 80% of chroma detail. Field notes: /field-notes/stillness, /field-notes/tripod-test, /field-notes/chroma-mtf, /field-notes/the-law-held-the-number-did-not, /field-notes/believing-instruments. Hugging Face collection: https://huggingface.co/collections/resoajoe/vibe-scientist-loglens-6a8b69d72dd4e8259c7c6605

Hypothesis

If generative video is used as a stand-in for a fixed camera (hospital room, site, listing walk), the footage will fail a stillness and chroma-integrity test long before a human notices, and any single collapse-time number from one model will not transfer to the next model.

Kill criteria

If a second independent architecture, measured on the same Orin protocol, holds stillness within 2x of the real camera (28.8 min) AND a chroma ablation that humans can see also moves SSIM below 0.95, the stillness/chroma story is architecture-specific folklore and this experiment is killed.

Validation criteria

The stillness gap remains >10x on at least two model families, the self-output collapse reproduces on a second family (even if the minute-mark moves), and SSIM (or the team's default quality gate) fails to flag a chroma loss that a stills-trained eye can see. Then publish the protocol, not a universal constant.

Signals observed

Real fixed camera 28.8 min still. Generated footage ≤4 s.

2026-08-01field note stillness

Destroy 80% of chroma; SSIM stays 0.9998 on three autoencoders on the Orin.

2026-08-01field note chroma-mtf

Same protocol vs a frontier model: 20 s vs 28.8 min (86x).

2026-08-10field note tripod-test

Collapse-on-self-output reproduced on a second architecture; the quoted death-point moved by ~5x.

2026-08-22field note the-law-held-the-number-did-not

Created: 2026-08-01

Updated: 2026-08-22