Move page polling off the Bot and onto a non-AI worker

How this helps you

Repeat "is it up yet" checks stop eating weekly usage, while you still get pinged when something real changes.

✦ Try it

I have a watchdog that keeps waking you to check [URL or page] for [what change matters]. Design a non-AI worker that polls on a [hourly] timer, writes a one-line result only when the change happens, and uses no model tokens while waiting. Show me the worker plan, where it will run on your computer, how I get alerted, and which Bot routine we should turn off after it works. Do not enable it overnight until I approve a one-day test.

What they did

A large share of Julian's usage advice is about not waking a full Bot to do dumb polling. He describes a pattern where the Bot writes a small worker program once. That worker has no AI in it, so it spends none of your Grok Bot tokens. It checks a page every hour (or whatever cadence you set). When something actually changes, the Bot gets a short result instead of rereading a long chat to ask "any update?". He contrasts this with watchdog routines that wake the Bot often and usually find nothing. The same idea shows up later with specialised scrapers (he mentions Apify bots) for map listings or comment collection: connect a tool that already does the job, then let Grok Bot summarise two lines. Commenters under the video were mostly fighting the soundtrack, but the polling tip is one of the few that is not already covered by a generic usage audit.

Who it's for: People whose routines mostly report "nothing new" and still burn the weekly cap.

Before you start

A Bot that can run scripts on its computer
One clear URL or page to watch
A definition of what counts as a change

How to try it

Pick one watchdog routine that mostly finds nothing (price page, listing, status URL).
Ask your Bot to write a small worker that fetches that URL on a timer and writes a one-line result file when something changes.
Confirm the worker runs without calling the model, then disable the old Bot routine that was polling.
Have the Bot read only the result file (or a short webhook) and message you when there is a real change.
Keep the first week of alerts on so you can catch a wrong selector before you trust it overnight.

What to expect

Instead of a Bot waking every hour to open a campsite page, a worker checks the page and only pings when a spot opens; the Bot then drafts the booking steps.

Common mistakes

Leaving the old Bot routine on in parallel with the worker.
Polling every few minutes after the usage cuts (Brock's commenters warn credits vanish by morning).
Skipping the one-day test and trusting a wrong selector overnight.
Asking the Bot to "keep an eye on it" in chat instead of writing a real worker.

Related tips

Have a Bot watch for cancellations at a fully booked spot: Ask a Bot to check a booking page every hour and ping you the moment a slot opens, so you can grab a cancellation before anyone else.
Turn a repeated job into a script the skill runs: Do a weekly job once with your Bot, check it, then have it rewrite the job as a plain script and save a skill that runs the script and reports in two lines.
Out of usage? Run these five checks before you buy credits: Grok Bot says you're out of usage but you still have credit? Check the weekly pool, the monthly cap, the meter, the account, then what's eating it.
Make your Bots read websites with a script, not screenshots: One test: 213K tokens by clicking around, 51K with a browser script, almost none with a plain script. Give your Bots the rule.