Ever rolled up to a ‘wheelchair accessible’ restaurant on Google Maps, only to face a flight of stairs and an awkward host offer to ‘carry you in’?
Yeah. That happened to a friend of ROLLIN’s CEO, David Sirota. Forty minutes wasted. And it’s not a glitch—it’s the entire model.
Checkbox accessibility. Yes or no. A binary lie flattening real-world mess into one database field. Most times, it’s blank or wrong. Tech loves it because it’s simple. Users hate it because life’s not.
That’s not accessibility. That’s a checkbox someone clicked in 2019 and never looked at again.
Sirota nails it there. I’ve seen this movie before—20 years covering Valley hype. Remember when Yelp promised crowdsourced truth, then drowned in fake reviews? Same vibe.
Why Does Checkbox Accessibility Suck So Bad for AI Agents?
AI’s coming. Fast. Claude’s MCP protocol, agents in your phone—they’ll book dinners, scout spots, without you lifting a finger. Pull from Maps API? Boom, same checkbox crap.
For able-bodied folks, wrong rec means meh pasta. For wheelchair users? Humiliation at the door. Or worse—stranded.
Here’s my unique take, one the original misses: this echoes the GPS data wars of the 2000s. Back then, free crowdsourced maps were a joke—bridges missing, turns into lakes. Then TomTom and Garmin built paid, verified layers. Drove the industry. Accessibility’s at that fork. Ignore it, and AI amplifies exclusion. Build it right, and it’s a goldmine.
Sirota’s dad has FSHD muscular dystrophy. Family pain fueled ROLLIN. Fair. But let’s ask: who’s cashing in? Consumer app’s free. Dev API? That’s where devs pay for scored JSON. Businesses embed widgets—hotels nudge guests to accessible eats. One data layer, multiple revenue streams. Smart, if it works.
What the Hell is ROLLIN, Anyway?
Not an app. Infrastructure. Scores 105k+ restaurants in 15 states—0-100 scale. Factors: ramps, restrooms, aisles, parking, elevators. Weighted. Verified.
Under the hood? Signal-processing engine. Merges sources, community checks, confidence scores. Won’t fake it—shows uncertainty instead. Better than pretending.
Three layers:
Consumer: Free web/iOS app. Agentic concierge—proactive or chill modes. On-device AI (Apple Vision) snaps photos, analyzes ramps, doors, lighting. No cloud. Privacy win.
Developer: REST API, Python SDK, MCP server on npm. Claude, Cursor, VS Code query natively. Ask for ‘wheelchair Italian near Midtown, good lighting’? Get JSON: scores, features, evidence.
Business: Widgets, badges. Same data.
Real scenario: Manhattan wheelchair user queries ‘accessible Italian near Times Square, no steps, good lighting.’ App spits ranked results—scores, breakdowns, sources. Tap one: entrance preview.
Sounds slick. But I’ve grilled founders like this. Early coverage? Hype heavy. Data coverage—15 states? US has 50. Confidence scores? Only as good as contributors. Community verification—pay-to-play risk?
Is ROLLIN Actually Better Than Google or Apple Maps?
Short answer: On paper, yes. Binary checkboxes are caveman tech. ROLLIN’s graduated, multi-source. But scale it.
Prediction: In two years, without this, AI travel agents become discrimination machines—subtly herding disabled users to bad spots. Regulators notice. Fines fly. ROLLIN (or rivals) wins by default.
Skeptical me wonders: Sirota’s promo screams ‘founder diary’—personal story tugs heartstrings. Classic Valley move. Weather data parallel? Spot-on, but weather’s commoditized free now. Will accessibility follow, crushing ROLLIN’s moat?
Dev angle—open source beat loves this. MCP integration means agents like Anthropic’s pull real intel. Python SDK auto-gen? Lazy dev dream. But who pays? Freemium tiers? Original mum.
Tested similar: Early APIs flopped on stale data. ROLLIN’s engine claims freshness—constant updates. Prove it.
Consumer side—photo AI’s killer. On-device means no creepy data hoover. Apple Vision crushes cloud rivals on privacy.
Business play: Restaurants badge-up for cred. Hotels embed—win-win? Or paywall for proof?
The Money Question: Who’s Getting Rich Here?
Always my opener. Sirota built for dad—not profit spin. But infrastructure? That’s VC catnip. Weather APIs rake millions yearly.
ROLLIN’s play: Power everything touching physical world. Calendars, Uber, TripAdvisor. If agents standardize on it—hello, monopoly rents.
Risk: Open data rebels. GitHub warriors scrape, fork, free it. Or Google buys ‘em out, checkboxes 2.0.
My bet—hybrid wins. Paid core, open edges. Devs flock.
Real Talk: Barriers Still Suck
Wheelchair entry? Sure. But FSHD? Muscle wasting hits arms, shoulders. Table height? Staff training? Unscored.
ROLLIN starts physical—good. Expand to conditions? Must.
Scenario deep-dive: That Manhattan query. Top hit: 92/100. Level entry (Street View + photo verify), wide aisles (floorplans), restroom (user report, 85% conf). Sources listed. Tap evidence—see pics.
Agent flow: Your Claude books it. Knows your profile—‘no steps, bright lights for low vision.’ Filters auto.
Game-changer? If data holds.
🧬 Related Insights
- Read more: AI Agents That Lie Less: A No-BS Framework for Self-Awareness
- Read more: Turso’s Rust SQLite Gambit: Edge Hype or Real Database Shift?
Frequently Asked Questions
What is ROLLIN accessibility platform?
ROLLIN’s an intelligence layer scoring physical accessibility (restaurants first) with APIs for AI agents and apps—no more binary checkboxes.
Will ROLLIN work with my AI assistant like Claude?
Yes—via MCP server and SDK. Queries return structured JSON with scores, features, confidence.
Is ROLLIN free for wheelchair users?
Consumer app’s free; devs/businesses pay for API/widgets.
