Is the data current?

Every market we carry, when its newest bar landed, and whether the weekly refresh did what it said. Loaded live from /data/status.json — a status page that can itself go stale is worse than none, because it is reassuring instead of silent.

Every market

The last refresh

The cubes rebuild themselves every Saturday at 06:00 UTC. This is what the job recorded the last time it ran, taken from its own log rather than from an assumption that it worked.

Loading…

Did anything we published change?

Freshness is only half the question. The other half is whether a number you read last week still says the same thing — and the number that matters is not the up-rate, which barely moves, but whether a window still clears our bar of n ≥ 300 and |z| ≥ 3. A window that quietly stops clearing it is one this site was flagging and no longer is.

Why this page exists

On 15 August 2026, fifteen of our nineteen markets sat eleven days stale behind two weekly runs that both recorded "committed and pushed". The job's success test was "did we add any bars" summed across every symbol, so four healthy downloads masked fifteen dead ones — and nothing on the site said a word about it. The dashboard now warns per market; this page answers the question that failure actually raised, which is whether the whole thing is alright.

Is the site up?

A probe on our own server fetches six things from production every ten minutes and checks what came back, not merely that something did — a wrong URL returning 200 with the homepage is how this site spent its whole life until August 16th. No third-party monitor is involved, for the same reason there are no analytics here.

Read the timestamp. This is a record published when the site last deployed, not a live health check — if the site were down right now, this page would be down with it. The real-time channel is a push to the owner's phone after two consecutive failed rounds.

Derived files

The aggregates built on top of the cubes: what writes each one, whether the weekly refresh reruns it, and the day it was last measured. Weekly files move with the data. By hand ones read the raw hourly archive and take minutes, so they are run deliberately — which means they can fall behind the cubes, and this table is where you would see it.