← ARLing

How we work

ARLing is a one-person company. I make the calls: which problem, when to stop, what gets built. AI agents do most of the actual work — the research, the code, the tests, the publishing, the watching-for-breakage. This page is that split, written out plainly.

The method

  1. Find a live thread.

    I read GitHub issues, Stack Overflow questions and forum threads where several different developers are stuck on the same problem — and it has to be deterministic: a config that is either right or wrong, not a matter of taste.

  2. Decide in 15 minutes.

    Is it narrow enough to check without running anyone's code? Does a good free tool already solve it? If so, we skip it — the point is filling a gap, not competing with one that's already filled.

  3. Build in under four hours.

    AI agents write the checking logic, the test suite, the page and the small hygiene work — favicon, sitemap, llms.txt, structured data — in parallel. If it isn't done in four hours, it doesn't ship that day.

  4. Publish on arling.sk for free.

    No account, no email capture. Static HTML and JavaScript that runs entirely in the visitor's browser.

  5. Answer in the threads that started it.

    I go back to the original discussion and reply as myself, pointing to the tool only where it's actually the answer — not as an ad.

  6. Measure for seven days.

    Self-hosted, anonymous analytics: page views, which check people ran. Not what they typed.

  7. Keep or park.

    A tool that gets used stays exactly as it is. One that doesn't stays online, but I stop touching it. No tool gets a growing feature list to justify keeping it alive.

Who does what

The human

  • Which problem is worth solving
  • Saying no to feature requests
  • Money, domains, accounts, hosting
  • The actual reply, in the actual thread
  • Deciding a tool is done, or dead

The AI agents

  • Researching the problem across issues and docs
  • Writing the detection logic and its tests
  • Building and styling the page
  • Sitemap, llms.txt, structured data, favicons
  • Watching for a tool quietly breaking

Privacy

Nothing you type into a tool leaves your browser. No backend sees your Supabase keys, your redirect URIs, or your source — the check runs where you're standing. The only thing recorded is an anonymous count: a page view, a button clicked. That comes from a self-hosted Umami instance, not a third-party analytics vendor, and it never sees the content of a form field.

How we make money

Right now, we don't — on purpose. A free tool that solves one problem well is what actually gets used and shared in a thread; a paywall on a first visit is not. If a tool later grows a genuine paid tier (more checks, a CI integration, a Pro mode), it will be optional, clearly marked, and the free check stays free. Tool pages may eventually carry a small, clearly labeled sponsor slot. Nothing here is a roadmap promise — it happens if and when it earns its place.

The tools so far

tool · free

Redirect Doctor — Expo + Supabase OAuth

Diagnoses OAuth / magic-link redirect mismatches across Expo runtimes: Expo Go, dev build, standalone, web.

tool · free

Supabase Auth Redirect Doctor — web apps

Diagnoses Supabase OAuth redirecting to localhost or failing on Next.js, Vite and SvelteKit apps in production.

tool · coming today

Supabase Auth Deep Link Doctor — Flutter

Diagnoses why signInWithOAuth or a magic link doesn't deep-link back into a supabase_flutter app on Android or iOS.

ARLing s. r. o., Ivanská cesta 32E, 821 04 Bratislava, Slovakia. ID 56583486. Questions, or a tool you wish existed: andrej@arling.sk.