Gloss

Open technical preview

A screenshot is not a PowerPoint.

Gloss measures whether generated decks look right and are built right — native shapes, charts, tables, layouts, text, and relationships included.

Inspect the repository Add a benchmark case

Generative Layout & OOXML Scoring System · Apache-2.0

submission.pptx
Q3 / SYSTEMS
The artifact
is the product.
Rendered pixels

108 visual checks inspect what reaches the canvas.

Gloss artifact X-raycanvas ↔ package
Evidence bundle Public JSON Open the receipts ↗
Candidate checks 280 108 visual · 172 inside-file
Positive controls 280/280 passed
Single-fault mutations 280/280 detected
Release state Preview 0 official model results

Frozen comparative v1

Four public baselines. Zero model claims.

Twelve editable decks, 240 graded slides, and three seeds per path. These are deterministic repository workflows built to expose the difference between visual similarity and native PowerPoint structure.

Inspect every deck and report Open summary JSON
comparative-v1 / summary.json 12 runs · 240 slides

Loading frozen baseline scores…

local artifact score; self-reported cohort loading

The 21-second version

The native gap, in 21 seconds.

The film is rendered from the frozen comparative summary above. These are repository-owned workflow baselines, not model scores.

00:211080 × 1080silentview renderer ↗

One file. Two truths.

Visual fidelity is necessary. Native structure is non-negotiable.

A generated deck can match the reference image and still collapse the moment someone edits a chart, applies a layout, changes a theme, or reads the file with another tool. Gloss grades both layers.

Read the open scoring contract
Rendered pixels108
The artifact twistinside the .pptx
Object structure101
Layouts & masters25
Package safety22
Text semantics20
Asset hashes4

Candidate checks by method. Counts come from preview-v1.json.

The repository is the product

Find a weak assumption. Open a patch.

No waitlist and no private benchmark committee. Cases, assertions, evidence, and release gates are inspectable on GitHub.

CASE

Break the corpus

Add the slide construct your generator gets wrong: charts, tables, masters, fields, typography, geometry, or package relationships.

Propose a benchmark case ↗ CODE

Make the grader stricter

Reproduce a false positive or false negative, then improve the assertion with a focused fixture and test.

Report a grader gap ↗ PROOF

Challenge the evidence

Question a provenance record, mutation operator, scoring assumption, or release gate. Every public claim should survive review.

Challenge a claim ↗
acidslide-v1/benchmark acidslide-v1/grader root/ACIDSLIDE_OPENSPEC.md root/CONTRIBUTING.md

Truth before leaderboard

Working harness. Unfrozen benchmark.

Gloss is ready for public collaboration, not official model ranking. The release validator fails closed until independent evidence and frozen artifacts exist.

✓
Candidate corpus validates

280 schema-valid checks and 381 traceable prompt clauses.

passing
✓
Mutation operators execute

280 controls pass; 280 generated single-fault mutations are detected.

passing
○
Independent evidence

Assertion provenance and evidence remain incomplete for 280 candidate checks.

open
✓
Comparative bundle

12 local runs, 240 slides, exact artifact hashes, and a one-command verifier are frozen.

passing
○
Official release

Independent prompt convergence, reviewer approvals, and signed release indexes remain open.

open

Reproduce it locally

Trust the command, not the claim.

Terminal
git clone https://github.com/aronchick/gloss.git
cd gloss
./acidslide-v1/benchmark/comparative-v1/reproduce.sh

Gloss is built in public

Make generated decks worth editing.

Join on GitHub