Sep 2024 to present · Europe

Aranx

Founding Engineer

Visit aranx.com ↗

I am building Aranx: search engine optimization and generative engine optimization on autopilot for founders and small businesses.

How it works

What it reads

  • Customer website
  • Competitor sites
  • Search data
  • Market trends

What it does

Aranx
  1. 1Research the market
  2. 2Pick a topic the site can win
  3. 3Write and illustrate the article
  4. 4Link related pages
  5. 5Publish

What comes out

A new article every day

Published on the customer's site, built to rank in search results and get cited in AI answers.

Articles generated

Loading article count

Why it exists

Buyers now ask AI assistants first.

Aranx keeps a small business visible in both places, search rankings and AI answers, without hiring a content team. It tracks the prompts buyers ask, records which brands the answer engines cite, and writes toward the gaps.

Under the hood

Built to publish without supervision.

I designed and built the whole engine. It runs serverless on Cloudflare's developer platform.

The agentic workflow

  1. 1

    Research

    Research agents study the customer's site and its competitors in real browser sessions.

  2. 2

    Keyword choice

    Planning agents pick topics the site can win using Ahrefs data: search volume, keyword difficulty, and what competitors already rank for.

  3. 3

    Draft

    Writing agents work from a brief and stay grounded in the site's own content through retrieval. Fast models classify, stronger models plan and write.

  4. 4

    Images

    Image models illustrate every article.

  5. 5

    Internal links

    New articles are linked into the rest of the site, in both directions.

  6. 6

    Review and publish

    Automated checks run, then the article ships on schedule.

Every article keeps an audit trail: why its topic was chosen, which sources grounded it, and which model wrote it.

The Cloudflare stack

Workers
The product surface and every API.
Agents on Durable Objects
One long-lived agent per customer site, holding context, memory, article queue, and schedule between runs.
Workflows
The per-article pipeline, one retryable step per stage. A failed step retries on its own and never stops the day's publishing.
Queues
Background jobs and fan-out, so publishing spikes never block the product.
Workers AI and AI Gateway
Model routing with caching, fallbacks, rate limits, and per-site spend caps.
AI Search and Vectorize
Retrieval over the site's own content, so drafts stay grounded in what the business actually does.
Browser Rendering
Real browser sessions for market and competitor research.
D1, R2, and KV
Article records and the audit trail, generated images, and configuration.