Baltor Showcase
baltor.ai

Every claim, with the checks behind it.

Baltor’s technical monographs pair each architectural claim with an executable verification suite, held-out counterfactual perturbation checks and known-wrong refusal controls.

Reproducible benchmarks, adversarial validation audits and pre-registered protocols. Four monographs, published October 2026.

How each claim is checked

Executable verification suite
Code that runs the claim’s measurement again.
Counterfactual perturbation checks
Altered cases, held out of development, that should move the result.
Known-wrong refusal controls
Inputs built to be wrong, which the method has to refuse.

Published verification monographs

October 2026, four monographs

  1. M01

    TAEDRI: composable code graphs vs. raw boilerplate codegen

    Evaluates about 19,000 parameterized node configurations and reports about 60% verified reuse, with lower repair latency.

    Code reuse
  2. M02

    Camouflaged alpha (α), international norms and Jev CPU partition routing

    Quantifies why LLMs get captured by surface SEC filings and narrow domestic statutes, and benchmarks a 4,738x search-space reduction from Jev probability routing.

    Routing and norms
  3. M03

    Adversarial benchmark validation (AV-AUC) and anti-benchmaxing audit

    Applies discriminative train and test shift classifiers and counterfactual AST rewrites across 24 industry benchmarks.

    Benchmark validity
  4. M04

    1,200-question quantitative UI/UX, TPSI, whitespace ratio and 2D saliency heatmaps

    Formalizes visual ergonomics, 4 pt and 8 pt harmonic padding and anti-AI-slop design rules, calibrated against paco.me, rauno.me and read.cv.

    Read M04

1,200-question quantitative UI/UX, TPSI, whitespace ratio and 2D saliency heatmaps

Monograph
M04
Published
October 2026
Publisher
Baltor
Calibrated against
3 reference sites

Abstract

M04 formalizes visual ergonomics as measures a page can be scored on. It sets out 1,200 quantitative questions about interface design, measures TPSI, whitespace ratio and two-dimensional saliency heatmaps, and states 4 pt and 8 pt harmonic padding and anti-AI-slop design rules. Its measures are calibrated against three reference sites: paco.me, rauno.me and read.cv [1, 2, 3].

The question set

The monograph scores an interface with 1,200 quantitative questions. Each question is answered with a measurement taken from the page itself.

Harmonic padding

Padding and gaps follow 4 pt and 8 pt steps, so spacing keeps one rhythm across the page.

4 pt
8 pt
12 pt
16 pt
24 pt
32 pt
48 pt
64 pt
Figure 1.A padding scale built from 4 pt and 8 pt steps, drawn at three times scale.

Whitespace ratio

The whitespace ratio compares the open ground of a layout with its whole area, which turns a sense of room into a number.

  • Content
  • Open ground
Figure 2.Whitespace ratio: the plain area as a share of the whole layout.

Saliency heatmaps

A two-dimensional saliency heatmap maps where attention lands on a page, so the headline and the main action can be checked against where the eye goes.

Figure 3.A saliency heatmap over the same layout, heaviest on the headline and the button.

Calibration

M04 calibrates its measures against three reference sites: paco.me [1], rauno.me [2] and read.cv [3].

Verification

Like every monograph at baltor.ai/showcase/papers [4], M04 pairs its claims with an executable verification suite, held-out counterfactual perturbation checks and known-wrong refusal controls.

References

  1. [1]paco.me, reference site for calibration. paco.me
  2. [2]rauno.me, reference site for calibration. rauno.me
  3. [3]read.cv, reference site for calibration. read.cv
  4. [4]Baltor. Published verification monographs, October 2026. baltor.ai/showcase/papers

Cite this monograph

Baltor. 1,200-question quantitative UI/UX, TPSI, whitespace ratio and 2D saliency heatmaps. Technical monograph M04, October 2026. baltor.ai/showcase/papers.

BibTeX
@techreport{baltor2026m04,
title = {1,200-Question Quantitative UI/UX, TPSI,
Whitespace Ratio and 2D Saliency Heatmaps},
institution = {Baltor},
type = {Technical monograph},
number = {M04},
year = {2026},
month = oct,
url = {/papers}
}

Save as PDF opens your browser’s print dialog with this page; Copy citation puts the BibTeX entry on your clipboard.