Noon is the second marketplace of the Gulf, and sellers on it work almost blind. There is no Helium 10, no Jungle Scout — no research layer at all. Sellers guess at demand, guess at price, and find out whether they were right after the inventory has already landed.

NoonSpy is the layer that was missing. It watches the marketplace, keeps a history nobody else keeps, and turns that history into a decision a seller can actually act on today.

The hardest constraint was not product — it was access. Noon sits behind Akamai, and a datacentre IP is refused on sight. Every conventional scraping architecture I tried died at the edge.

So the architecture had to invert. Instead of the server reaching out for data, the data reaches in: a collector runs from a residential connection and pushes snapshots to an authenticated endpoint on the platform. The server never scrapes. It only receives, normalises and reasons.

One artefact carries the whole product: the Verdict. A seller pastes any Noon URL and gets back Sell / Risky / Skip, five component scores, and a written next action.

Everything else — trending, niches, price history, alerts — exists to make that single answer better. Free accounts see the scores. Paid accounts see the reasoning. That line is the business model, and it is enforced in exactly one function so it can never drift.

A dense product for people who read numbers all day. Data-first typography, one accent colour reserved for the verdict itself, and no decoration competing with the signal.

The interface had to survive being open for eight hours. That ruled out high-contrast dashboards and pushed everything toward calm surfaces with sharp, deliberate accents at the exact points a decision is made.

Python and Flask on a VPS, SQLite in WAL mode after a concurrency bug taught me why, Stripe for subscriptions with per-charge statement descriptors, and a Manifest V3 Chrome extension that injects live NoonSpy data straight into the Noon product page.

The AI layer runs on the Claude API — the Verdict reasoning, the listing copy generator, and an image-pack studio for seller creatives. Each one is metered against the same access check.

A working, billing, self-serve SaaS with a live user base — built and maintained solo, from the scrape architecture up to the pricing page.

It is also the sharpest thing on my desk: every feature had to justify itself against a real seller's next hour of work.