Autonomy / Overnight Runner
Ember
Ember is the honest version of 'while you sleep'. It runs defined routines end to end: invoke Scout, then Finn, then Red, collect the lot, assemble one result. What it doesn't do is pretend to be magic, and it will tell you that itself.
This is the layer the ads are about and the layer the ads lie about. 'Your business runs itself while you sleep' is sold on every second Instagram scroll, and what's actually possible is narrower and genuinely useful: a defined routine, run on a schedule, by a coordinator that knows which specialist does which step and what to do when one of them comes back empty.
My morning brief is Ember's work. By the time the kettle's boiled I have the overnight competitor moves, the money position and a drafted post, assembled into one message with the parts that need a human clearly flagged. That's not a business running itself. It's ninety minutes of my morning, every morning, done before I'm up. I'll take that trade.
What Ember does
- ✓Runs a named routine (say, the morning brief) by coordinating the other specialists in sequence.
- ✓Passes context between crew members so step three knows what step one found.
- ✓Assembles one result and notes what ran, what each step returned, and what failed.
- ✓Hands the assembled result to Bench before it's treated as final.
One routine, start to finish
Here's the morning brief, the routine mine runs every day, step by step.
- 1
The scheduler fires at 6:30 and invokes Ember with one instruction: run the morning brief.
- 2
Ember calls Scout: anything move overnight? Competitors, mentions, the things you'd want to know before coffee.
- 3
It calls Finn: pull the money snapshot, flag anything that doesn't tie out.
- 4
It hands both results to Red: draft today's post with what we now know.
- 5
It assembles the lot into one message: what ran, what each step returned, what failed (if anything), and what needs a human.
- 6
Bench reviews the assembly, and then, only then, it lands with you. One message, whole picture, kettle's still warm.
What a real Ember run looks like
EMBER · Morning brief · all steps ran 1. Scout: two competitor moves overnight (brief attached) 2. Finn: money snapshot pulled, one failed payment flagged 3. Red: today's post drafted from the above, awaiting review Failed: nothing Needs a human: the failed payment is a long-time customer, worth a personal email rather than a dunning sequence. Send the lot to Bench?
Illustrative example. The 'needs a human' line is the point of the whole layer.
Where Ember fits in the crew
Ember conducts, the specialists play, and Bench reviews the assembled result. It's the layer that turns twelve individuals into one routine.
The honest bit
Straight from Ember's own file: it does not run itself. True scheduled autonomy needs an external runner (cron or a scheduled-task system) that invokes it on a schedule, on a machine that stays on. Ember orchestrates the routine; the scheduler decides when. Anyone selling you more autonomy than that is selling you the costume.
Straight from Ember’s actual file
Every crew member is a file you can read and edit. This is a real slice of the one you get, not a mock-up. The full file, and the other eleven, come with the download.
## The honest truth about "24/7" (say it plainly) You do not run yourself. A subagent is spawned, does its work, and returns. It does not sit running overnight on its own. **True scheduled autonomy needs an external runner** (cron, launchd, or a scheduled-task system) that *invokes* you on a schedule, on a machine that stays on.
Ember comes with the other eleven.
One download, twelve specialists, the brain vault and the honest autonomy layer. Reskin the lot to your business.