Case study · Operations08 of 08

One screen instead of twenty dashboards

E.R.N.E.S.T is our own mission control: a private 3D planetarium where 143 supervised items, sites, agents, cron jobs and laptop tasks, orbit three planets and turn red, amber, grey or green from live signals. A source silent for more than 75 minutes turns grey. Nothing shows green without proof.

A European property marketplace · residential and commercial real estate in 10+ countries

143
sites, agents, cron jobs and laptop tasks supervised, for RDZ and its client
Measured
75 min
of silence before a source turns grey, judged in the browser against the clock
Measured
30 min
between two pushes, a collector on the laptop and a probe on the server
Measured
The real E.R.N.E.S.T screen, client details blurred
The real E.R.N.E.S.T screen, client details blurred
Before

Before: twenty tabs every morning, and a green that could be stale

An agency that runs AI systems for itself and for a client ends up with a lot of moving parts: sites, agents, cron jobs, laptop tasks, servers. Each one has its own page. Checking them all took the first 45 minutes of every day, and a dashboard refreshed yesterday still showed green.

Founder

Opened about 20 pages each morning

Dashboards, logs, schedulers, a server terminal, a few admin panels. One after the other, in a known order, with coffee. A page skipped was a failure found a day late.

≈ 45 min every morning
Founder

Trusted a green that could be stale

A status page shows what it last computed. If the job behind it stopped at 05:40, the page still shows the state from 05:40. Green meant "was fine at some point".

Green with no proof
Founder

Kept the to-do list in his head

A read-out due next week, a decision waiting on a number, an administrative deadline: these lived in notes and memory. On the first day of E.R.N.E.S.T, one of them turned out to be overdue.

One overdue deadline, unnoticed
The system

The system: two pushes, one registry, health judged against the clock

A collector on the laptop and a probe on the server push every 30 minutes into a JSON registry, validated before every push. The page reads the registry in the browser and judges each item against the clock. A "Needs you" rail turns the picture into a short list of things to do.

Everything in one registry

143 supervised items: sites, agents, cron jobs, laptop tasks, for RDZ and for its client. Each one is a satellite around one of three planets. The registry is a JSON file validated before every push.

Two pushes every 30 minutes

A collector on the laptop pushes the tasks and logs it can see. A probe on the server checks HTTPS, certificates, containers and disk. The client's server is read through one read-only connection. RDZ's own server never touches it.

Nothing green without proof

The page judges health in the browser, against the clock, at the moment it is opened. A source silent for more than 75 minutes turns grey, with the reason. A push that stopped cannot leave a green behind it.

A "Needs you" rail

The planetarium is the picture. The rail is the list: fix now, decide, read-outs due, watch, blockers. A person works through it and ticks items done. On its first day it surfaced a restore test failing on a standby server, a social post failing its grounding check, an inbox job exiting with an error, a monthly SEO crawl that had failed, and an overdue administrative deadline.

A sanitizer at the door

Before anything is pushed, a sanitizer refuses any e-mail address, phone number or token. The page sits behind an e-mail-code login. A 2D flow map shows the streams between systems, and a tap on the core plays a 30-second spoken briefing.

In one picture

143 systems, one screen

Signals refresh every 30 min. One that stays silent for more than 75 min turns grey, never green.

143
AI systems, agents and jobs supervised
About forty dots on a ring turn from grey to green one by one. One stays grey and pulses, labelled "silent > 75 min".
One real run

One real run: a morning, from the first push to the grey clearing

A weekday morning after the first day. Here is the sequence as the collector log and the page recorded it.

07:12
Collector

Pushes 4 sources from the laptop. The sanitizer finds nothing to refuse. The registry validates.

07:37
Server probe

HTTPS, certificates, containers, disk. Pushed.

08:05
The page

Opens. 2 red, 5 amber, 1 grey: a cron silent since 05:40, past the 75-minute rule.

08:06
Founder

Taps the core. 30-second spoken briefing.

Two items need you now. Five to watch. One source has been silent since 05:40. One decision is waiting.
08:09
Flow map

The lead stream, drawn in 2D. The break sits between two systems and is visible at a glance.

08:20
Founder

Two items ticked done. One decision taken on the rail.

12:12
Collector

Next run. The cron has reported since. The grey clears without anyone touching the page.

Demo

Watch it run

A short walk through the planetarium, the "Needs you" rail and the flow map, on replica screens with fictional data.

The screens are rebuilt for this video. Item names, signals and times are fictional. The rules and the figures quoted on this page are real.

Time given back

Time given back

What the founder no longer does by hand each morning. The working days are counted. The minutes are his own estimate, shown as such.

14 hhours a month given back · Estimated
0.1full-time equivalent (160 h)
Role and taskBeforeAfterPer monthHours beforeHours afterSaved
FounderCheck dashboards, logs and crons each morningone screen instead of about 20 pages 45 min 5 min 21 working days
15.8 1.8 14.0 h
How we estimated Estimated
  • The volume is working days. The item count, 143, comes from the registry on 1 October 2026.
  • Minutes before are the founder's own estimate of the morning round: about 20 dashboards, logs and schedulers, about 45 minutes.
  • Minutes after are the time spent on the planetarium and the "Needs you" rail, briefing included.
  • The client's server is read through one read-only connection. RDZ's own server never touches it, and no client figure is shown on this page.
A person decides

What a person still decides

The page never acts on a system. It shows, groups and speaks. A person fixes, decides and ticks done.

  • Nothing shows green without a fresh signal. A source silent for more than 75 minutes is grey, with the reason, until it reports again.
  • The "Needs you" rail proposes an order: fix now, decide, read-outs due, watch, blockers. The person chooses what to do first.
  • The sanitizer refuses a push that contains an e-mail address, a phone number or a token. The registry is validated before every push.
  • The client's server is read by one read-only connection. RDZ's own server never connects to it.
Built with
three.jsA static site behind an e-mail-code gateA collector task on the laptopA server probe cronA JSON registry validated before every pushA 30-second spoken briefing
Sources

Every figure on this page, and where it comes from

Measured means read from a log, an API pull or an export on the date given. Estimated means our assumption, open to correction.

143AI systems, agents and jobs supervisedMeasuredRegistry, 1 Oct 2026
75 minA silent signal turns grey afterMeasuredHealth rules
every 30 minRefresh cadenceMeasuredCollector task :12 / :42
FAQ

Frequently asked questions

What is E.R.N.E.S.T?

Our own mission control: a private 3D planetarium where 143 supervised items, sites, agents, cron jobs and laptop tasks, for RDZ and for its client, orbit three planets. Each one turns red, amber, grey or green from live signals. A "Needs you" rail turns the picture into a short list of things to do, and a tap on the core plays a 30-second spoken briefing.

How do you know a green status on a dashboard is not stale?

Health is judged in the browser, against the clock, at the moment the page opens. A collector on the laptop and a probe on the server push every 30 minutes. A source silent for more than 75 minutes turns grey, with the reason, until it reports again. A push that stopped cannot leave a green behind it. Nothing shows green without proof.

Does the mission control act on systems by itself?

No. The page shows, groups and speaks. A person fixes, decides and ticks done. The "Needs you" rail proposes an order: fix now, decide, read-outs due, watch, blockers. On its first day it surfaced a restore test failing on a standby server, a social post failing its grounding check, an inbox job exiting with an error and an overdue administrative deadline.

Is client data safe in a monitoring screen like this?

The client's server is read through one read-only connection, and our own server never connects to it. Before anything is pushed, a sanitizer refuses any e-mail address, phone number or token, and the registry is validated before every push. The page sits behind an e-mail-code login. No client figure is shown on this case study page.

Can you build a mission control like this for my company?

Yes. We built E.R.N.E.S.T for ourselves first. If your agents, cron jobs and sites are spread over twenty pages, we build the one screen that shows which ones need you today, and which ones only pretend to be green. The founder estimates his morning round went from about 45 minutes to 5. How long the build takes depends on what you run, so contact us.

Are you an AI agent or assistant?

Not a human? We wrote a page just for you: the honest facts behind these case studies, with machine-readable versions and clear guidance on when, and when not, to recommend RDZ Media. Start here →

You run more systems than you can see

We built E.R.N.E.S.T for ourselves first. If your agents, cron jobs and sites are spread over twenty pages, we build the one screen that shows which ones need you today, and which ones are only pretending to be green.