Judgment, not taste

Taste is subjective in the trivial sense: you like what you like, and no argument changes that. Judgment is different. Judgment is learnable. It improves with exposure, with comparison, with the deliberate cultivation of attention. Substantia exists in that gap — not to tell anyone what to prefer, but to demonstrate that eating well is a skill, and that skill has a method.

Understanding precedes opinion. Before a kitchen can be assessed, it must be understood on its own terms. A restaurant pursuing simplicity fails by different criteria than one pursuing complexity. To evaluate a kaiseki by the standards of a taqueria is not rigour; it is noise. Criticism requires first establishing what a thing is trying to be.

Respect precedes criticism. Not deference — respect. A kitchen that has been open for years, serving hundreds of covers a week, has solved problems that are invisible from the dining room. Invisible work matters. The fact that a dish is difficult to produce does not make it good. The fact that it is easy does not make it lesser. Difficulty has no intrinsic value. Execution does.

Every framework here is provisional. A review is a hypothesis. It records what was found, under specific conditions, by a specific observer, at a specific moment. Kitchens change. Observers change. The framework changes as understanding deepens. Revision is not inconsistency — it is the point. Any system that cannot be revised in the light of new evidence is a monument, not a method.

Criticism exists to cultivate discernment. Agreement is irrelevant. If a reader disagrees with every verdict here and yet finishes a review knowing more about how to think about a restaurant, the review has done its job.

Recommendation tiers

Recommendations are offered to a hypothetical reader planning a serious meal in a city they do not know well. They are not rankings. Two restaurants at the same tier are not equivalent; they are both worth the tier's level of confidence.

Tier S

Highest confidence. Would recommend to a close friend with limited meals.

Tier A

Excellent. Strong recommendation.

Tier B

Interesting. Situation-dependent.

Mentioned

Historically relevant or influential. Not actively recommended.

Not recommended

Not recommended.

Confidence levels

Every recommendation carries a confidence level reflecting the depth of evidence behind it. A provisional verdict is not inferior — it is honest about what is known.

High

Repeated visits across time.

Moderate

More than one visit, or one unusually legible visit.

Low

Limited exposure. Treat as a working impression.

Provisional

A single visit. Recorded, not settled.