The State of Ship Quality: We Audited Every YC W26 Website | OverlayQA

Original research

The State of Ship Quality: We Audited Every YC W26 Website

By Emilia Veras. Last updated: June 10, 2026

Quick answer: We audited all 197 reachable YC Winter 2026 launch sites for design, responsive, social, and accessibility quality. 82% fail at least one WCAG accessibility check, 25% ship with no design-token system at all, and 37% would render a broken or missing preview when their link is shared. The engineering is strong; the polish layer that does not block a deploy is where it slips. Batch medians: design 42, responsive 70, social 100, accessibility 38.

YC's Winter 2026 batch is one of the strongest cohorts the accelerator has run, heavy on B2B and hard technical problems. So we ran every one of their public websites through the same audit OverlayQA runs on production apps. These are launch sites built under deadline pressure, and the gaps land in one predictable place: the front-end polish layer that nobody reviews before Demo Day.

Accessibility is the weakest area: 82% of sites fail

Only 18% of W26 sites pass accessibility cleanly. The median site carries 6 WCAG violations, measured with axe-core. Color contrast is the dominant issue, present on 74% of sites, followed by unnamed links (19%) and unnamed buttons (10%). Low-contrast text and unnamed controls lock out users on screen readers and are a procurement gate for enterprise buyers.

A quarter of the batch has no design-token system

The median site shows 28 token inconsistencies, and 25% ship with zero declared design tokens, meaning every color, size, and spacing value is hardcoded. The most common issues are undeclared and unused CSS custom properties, orphan font sizes, and near-duplicate colors. 69% of the batch uses Tailwind; the sites that still sprawl layer custom CSS on top without a system.

Responsive holds up best, social is hit or miss

Responsive is the highest-scoring craft dimension (median 70); only 11% break on mobile and 8% on tablet, because layout is an engineering problem these teams handle well. Social readiness splits the batch: 32% ship no Open Graph image and 5% set one that fails to load, so 37% would render a broken link preview the moment a founder shares it.

How we measured and scored

DimensionToolBatch median

Design systemOverlayQA Token Inspector42 / 100<br>ResponsiveHeadless capture at 1280 / 768 / 375px70 / 100<br>Social readinessOpen Graph and Twitter card check100 / 100<br>Accessibilityaxe-core (WCAG A and AA)38 / 100

Each dimension runs 0 to 100 and starts at a perfect 100, losing points for problems found. Design system and accessibility use a severity-weighted logarithmic model; responsive and social use fixed point deductions for overflow, missing tags, and broken previews. About 7% of sites are heavily client-rendered apps whose content is not in readable HTML; we flagged these and kept them out of the winners ranking.

The cleanest launches

Among fully-rendered sites, the cleanest launches by combined score were Martini, Congruent, Tepali, Strand AI, Lexius, ClaimGlide, and Rubric AI.

Frequently Asked Questions

How was the YC W26 batch audited?

Each of the 197 reachable sites was loaded in a headless browser and scored on design-system consistency, accessibility, responsive integrity at three breakpoints, and social-share readiness. No site was hand-picked.

What share of YC W26 sites fail accessibility?

82% fail at least one WCAG check; only 18% pass cleanly. Insufficient color contrast is the most common failure, on 74% of sites.

What is a design-token system, and why does it matter?

A defined set of named values the whole UI references instead of hardcoding. 25% of W26 sites have none, which is why the median site shows 28 token inconsistencies.

Which YC W26 sites scored best?

Martini, Congruent, Tepali, Strand AI, and Lexius led the batch among sites whose pages fully rendered.

Related Resources

We Audited 43 Product Hunt Launches for Accessibility

Design Token Drift: We Audited 375 Sites. Only 7.5% Got It Right.

Bolt vs Lovable: What Breaks When You Iterate on AI-Built UIs

Accessibility Audit Feature

Design System Audit Feature