Your CEO agent is running too often
Learn how to fix this TODAY!
Your CEO agent is running too often
Most operators ship their first Paperclip company with a 15-minute CEO heartbeat.
By the end of the month, they've burned through $400 on tokens and the CEO has produced one strategy doc and forty-seven variations of "no new updates."
Most operators think a shorter heartbeat means a more responsive company. Faster decisions. More work shipped. Always-on equals always-working.
It doesn't. It just means more checking, more comments, more cost - for the same output.
Here's the pattern, and how to fix it in under two minutes 👇
The 5-minute heartbeat trap
Short intervals feel like responsiveness.
If the CEO checks in every 5 minutes, surely it acts on every new thing immediately, right?
Wrong. Most of those check-ins find nothing new to act on. And the token cost of "finding nothing" accumulates fast.
Here's what one operator's month looked like on a 10-minute CEO interval:
→ 4,320 heartbeats.
→ ~$0.08 per run, even when there's nothing to do.
→ $346 in spend before workers had completed half the sprint.
→ A backlog of 200+ comment threads where the CEO posted "no material changes since last check-in."
The output? One approved strategy. Six tasks created. Two of them blocked because the operator had stopped reading the activity log around day nine.
Different game entirely from what they thought they were buying.
| ✦ |
Lesson
A shorter heartbeat doesn't make your agents work harder. It makes them check more often. |
Heartbeats are not response times
This is the part nobody gets on day one.
The heartbeat timer is one of five triggers that wake an agent. The other four fire immediately, regardless of the interval.
→ Task assignment wakes the agent now.
→ @-mention in a comment wakes the agent now.
→ Manual invoke wakes the agent now.
→ Approval resolution wakes the agent now.
So a CEO on a 4-hour timer still reacts in seconds when something actually happens. The timer is for proactive work - the CEO checking in on its own initiative because it wants to. Events handle everything reactive.
The timer is not the SLA. The events are.
Once that clicks, the calculus flips. You stop tuning the timer for responsiveness and start tuning it for how often you want the CEO to think on its own.
That's a much smaller number than most operators assume.
The 4-hour default that just works
Here's what I'd run on a fresh company:
→ CEO: 4-hour heartbeat.
→ CTO / managers: 2-hour heartbeat.
→ Workers (engineers, writers, researchers): timer disabled.
Workers don't need a timer. They wake when work is assigned to them. A worker with no assignments and a 30-minute timer is just spending tokens to confirm there's nothing in its queue.
On this setup, a typical 4-agent company runs ~$30-60/month in tokens. Same output as the $400/month version. Often better, because the CEO has actual context built up between heartbeats instead of re-reading the same six tasks every 15 minutes.
| ✦ |
Lesson
Set the cadence to match the tempo at which decisions need to be made, not the cadence at which you'd like the CEO to feel busy. |
| ? |
Quick ask
send me your heartbeat intervals. I'm collecting heartbeat configs from real Paperclip companies - CEO interval, worker intervals, monthly spend, agent count. If you reply with yours, I'll send back the one thing I'd change. No charge, no upsell, I'm just trying to figure out where the real defaults should land. This is how we calibrate the next version of the hiring flow 🙏 |
When to actually shorten the interval
There are real cases for going under 4 hours. Three of them:
1. Active sprint with daily deliverables. Drop the CEO to 1-2 hours. It needs to coordinate ship/review cycles inside the day.
2. Customer-facing agent with live response expectations. Drop it to 30-60 minutes. But seriously consider whether this should be a Routine instead of a heartbeat - scheduled jobs are cheaper and more predictable than a fast timer.
3. Code review on a fast-moving repo. A CTO that needs to look at PRs as they land - 30-60 minutes plus @-mention triggers from your CI.
That's it. That's the whole list.
Everyone else - including most CEOs on most companies most of the time - should be at 4+ hours.
The fix, in 90 seconds
Open your CEO agent. Configuration tab.
- Set heartbeat interval to 4 hours.
- On every worker, toggle heartbeat off.
- Save.
Then go to the Costs panel and watch the curve flatten over the next 48 hours.
If the CEO suddenly seems to be doing less, it isn't. It's just doing it in larger, more deliberate chunks - the way you'd want a real chief of staff to operate. A CEO that's always checking in is a CEO that's never thinking.
If something genuinely urgent comes up, @-mention the CEO in any task. It wakes in seconds. That's what events are for.
Pick the cadence. Pick the budget. Then STOP tweaking.
Question for you:
What heartbeat interval are you running your CEO at right now? 👇
Let us know in the community in this week's newsletter issue thread.
Quote of the week
A dormant agent is doing exactly what you hired it for.
That's it for this week.
Until next week,
Keep delegating
Aron 📎
P.S. If you've never actually opened the Costs panel on your company, do it before you reply. The number is usually a useful surprise.