ReadEasy

Inspiration

Great startup ideas die when dense pages block real life: a student missing a financial-aid deadline, an applicant misreading admissions guidance, a family misunderstanding benefits instructions. Around 15–20% of people are neurodivergent, and remediation vendors charge tens of thousands per site audit while content goes stale the next day. I saw a startup-shaped gap: readers need a free tool now, and publishers need an affordable way to stay compliant with ADA, Section 508, and EU accessibility expectations.

What it does

ReadEasy takes a pasted URL or raw text and returns plain language with an action checklist. The demo trio — IRS benefits, UT Dallas admissions, and USCIS-style guidance — shows the same flow a paying customer would use: messy source page in, structured readable page out. Readers choose Focus, Dyslexia with Bionic reading support, ADHD micro-cards with one idea per screen, Listen with karaoke highlighting, or the Action checklist that turns deadlines into steps, plus Ask this page for grounded questions, reading-level choices, cosmetic toggles, history, and export. The same pipeline doubles as the business: one API endpoint that renders an accessible version of any publisher page.

Try it: https://readeasy-founders.vercel.app — Code: https://github.com/GhostInHex/readeasy/tree/hackathon/next-founders

How I built it

The pipeline is Fetch to Clean to Restructure to Render. Fetch retrieves the page, Clean strips ads, navigation, and scripts with Mozilla Readability plus jsdom without AI, Restructure calls OpenRouter under a strict JSON contract with title, summary, reading time, action items, and sections with key takeaways, and Render draws that JSON through a mode registry in Next.js 15 with React and TypeScript. A three-model fallback chain, one retry on malformed JSON, and structured errors keep it reliable, and the stateless design with no database or login deploys cleanly on Vercel.

Challenges I ran into

Some government pages block server fetching with a 403, so I built a raw-text paste path that runs the identical pipeline and cached the demo trio so the pitch never depends on live network. The no-hallucination constraint shaped the architecture: the model reshapes only text already present, deadlines are copied verbatim, and Ask this page answers strictly from page content. That discipline became the selling point — a verifiable transform instead of a summary judges cannot trust.

What I learned

I learned to pitch architecture as economics: each transform costs fractions of a cent in model usage against subscription and per-transform API pricing, which supports 75–88% gross margins. The numbers behind the plan are a three-year obtainable market of roughly $2–6M ARR from university Team plans and enterprise API customers, with a break-even near $168k ARR.

What's next

The roadmap is freemium readers today, Pro at $9 per month and Team at $49 per month next, then the enterprise compliance API at volume-tiered per-transform pricing with LMS embeds and a white-label option — the free student surface as distribution, the API as revenue, scaling horizontally on serverless infrastructure.

Built With: Next.js 15, React, TypeScript, Mozilla Readability, jsdom, OpenRouter, Vercel.

Provenance

Built on a ReadEasy base from the Buildverse window Sep 1–6, extended with the business model, architecture docs, and pitch framing in this branch for Next Founders.

Built With

• jsdom
• mozilla-readability
• next.js-15
• openrouter
• react
• typescript
• vercel